Revista Facultad de Ingeniería, Universidad de Antioquia, No.111, pp. 64-75, Apr-Jun 2024
ADHE: A tool to characterize higher education
dropout phenomenon
ADHE: Una herramienta para caracterizar el fenómeno de deserción en educación superior
Oscar Daniel Rivera-Baena1, Carmen Elena Patiño-Rodríguez1*, Olga Cecilia Úsuga-Manco1,
Freddy Hernández-Barajas2
1Departamento de Ingeniería Industrial, Universidad de Antioquia. Calle 67 # 5 3-108. C. P. 050010. Medellín, Colombia.
2Escuela de Estadística, Facultad de Ciencias, Universidad Nacional de Colombia. C. P. 050034. Medellín, Colombia.
CITE THIS ARTICLE AS:
O. D Rivera-Baena, C. E.
Patiño-Rodríguez, O. C.
Úsuga-Manco and F.
Hernández-Barajas. ”ADHE: A
tool to characterize higher
education dropout
phenomenon”, Revista Facultad
de Ingeniería Universidad de
Antioquia, no. 111, pp. 64-75,
Apr-Jun 2024. [Online].
Available: https:
//www.doi.org/10.17533/
udea.redin.20230519
ARTICLE INFO:
Received: June 16, 2022
Accepted: May 02, 2023
Available online: May 02, 2023
KEYWORDS:
Student dropout proportion;
academic analytic; dashboard;
Data visualization; higher
education
deserción escolar; analítica
académica; dashboard;
visualización de datos;
enseñanza superior
ABSTRACT: The field of academic analytics emerged in higher education institutions
(HEI) because of developments in database technologies and the generalization
of data-mining practices and business intelligence tools. We have designed and
implemented a dashboard called ADHE (Academic Analytical Dashboard in Higher
Education) for a Colombian higher education institution. The purpose of ADHE is to
help administrators of academic programs in their decision-making process regarding
the analysis of the phenomenon of student dropout. We used the pipeline methodology
for processing large volumes of data was used, which is based on five phases: data
acquisition, integration, cleaning, transformation, and visualization. All phases were
carried out in the R programming language using academic information sources from
the Faculty of Engineering of the Universidad de Antioquia and the Colombian Institute
for the Evaluation of Education. The dashboard ADHE is open for free and can be
consulted at: https://fhernanb.shinyapps.io/AppPermanencia/. The main findings were
that social stratum, gender, and type of high school are associated with the student
dropout phenomenon. Furthermore, in social stratum 1, male students and public high
schools tend to have a higher student dropout proportion. Additionally, we conclude that
admission to engineering programs requires a balance of qualitative and quantitative
abilities. The dashboard ADHE should be used to help students, parents, teachers, and
administrators understand student dropout dynamics.
RESUMEN: El área de analítica académica emergió en instituciones de educación superior
por causa del desarrollo de la tecnología en la recolección de información. En este
trabajo se presenta el diseño e implementación de un Dashboard de analítica académica
en una institución de educación superior de Colombia para apoyar el proceso de
toma de decisiones de los administradores de programas académicos con relación
a la deserción. Se empleó la metodología pipeline de procesamiento de grandes
volúmenes de datos la cual está basada en cinco fases: adquisición, integración,
limpieza, transformación y visualización de datos. Todas las fases se llevaron a cabo
en el lenguaje de programación R utilizando fuentes de información académica de
la Facultad de Ingeniería de la Universidad de Antioquia y del Instituto Colombiano
para la Evaluación de la Educación. El dashboard ADHE es de acceso libre y se
puede consultar en https://fhernanb.shinyapps.io/AppPermanencia/. Los principales
resultados obtenidos fueron que el estrato socioeconómico, el género y el tipo de colegio
están asociados con el fenómeno de deserción. Se encontró que el estrato social 1, los
estudiantes masculinos y el tipo de colegio público tienen las proporciones de deserción
más altas.
El dashboard ADHE podría ser usado por estudiantes,
padres de familia, profesores y administradores para
entender la dinámica de la deserción estudiantil.
64
* Corresponding author: Carmen Elena Patiño-Rodríguez
E-mail: elena.patino@udea.edu.co
ISSN 0120-6230
e-ISSN 2422-2844
DOI: 10.17533/udea.redin.20230519
64
ADHE: A tool to characterize higher education
dropout phenomenon
ADHE: Una herramienta para caracterizar el fenómeno de deserción en educación superior
Oscar Daniel Rivera-Baena1, Carmen Elena Patiño-Rodríguez1*, Olga Cecilia Úsuga-Manco1,
Freddy Hernández-Barajas2
1Departamento de Ingeniería Industrial, Universidad de Antioquia. Calle 67 # 5 3-108. C. P. 050010. Medellín, Colombia.
2Escuela de Estadística, Facultad de Ciencias, Universidad Nacional de Colombia. C. P. 050034. Medellín, Colombia.
CITE THIS ARTICLE AS:
O. D Rivera-Baena, C. E.
Patiño-Rodríguez, O. C.
Úsuga-Manco and F.
Hernández-Barajas. ”ADHE: A
tool to characterize higher
education dropout
phenomenon”, Revista Facultad
de Ingeniería Universidad de
Antioquia, no. 111, pp. 64-75,
Apr-Jun 2024. [Online].
Available: https:
//www.doi.org/10.17533/
udea.redin.20230519
ARTICLE INFO:
Received: June 16, 2022
Accepted: May 02, 2023
Available online: May 02, 2023
KEYWORDS:
Student dropout proportion;
academic analytic; dashboard;
Data visualization; higher
education
deserción escolar; analítica
académica; dashboard;
visualización de datos;
enseñanza superior
ABSTRACT: The field of academic analytics emerged in higher education institutions
(HEI) because of developments in database technologies and the generalization
of data-mining practices and business intelligence tools. We have designed and
implemented a dashboard called ADHE (Academic Analytical Dashboard in Higher
Education) for a Colombian higher education institution. The purpose of ADHE is to
help administrators of academic programs in their decision-making process regarding
the analysis of the phenomenon of student dropout. We used the pipeline methodology
for processing large volumes of data was used, which is based on five phases: data
acquisition, integration, cleaning, transformation, and visualization. All phases were
carried out in the R programming language using academic information sources from
the Faculty of Engineering of the Universidad de Antioquia and the Colombian Institute
for the Evaluation of Education. The dashboard ADHE is open for free and can be
consulted at: https://fhernanb.shinyapps.io/AppPermanencia/. The main findings were
that social stratum, gender, and type of high school are associated with the student
dropout phenomenon. Furthermore, in social stratum 1, male students and public high
schools tend to have a higher student dropout proportion. Additionally, we conclude that
admission to engineering programs requires a balance of qualitative and quantitative
abilities. The dashboard ADHE should be used to help students, parents, teachers, and
administrators understand student dropout dynamics.
RESUMEN: El área de analítica académica emergió en instituciones de educación superior
por causa del desarrollo de la tecnología en la recolección de información. En este
trabajo se presenta el diseño e implementación de un Dashboard de analítica académica
en una institución de educación superior de Colombia para apoyar el proceso de
toma de decisiones de los administradores de programas académicos con relación
a la deserción. Se empleó la metodología pipeline de procesamiento de grandes
volúmenes de datos la cual está basada en cinco fases: adquisición, integración,
limpieza, transformación y visualización de datos. Todas las fases se llevaron a cabo
en el lenguaje de programación R utilizando fuentes de información académica de
la Facultad de Ingeniería de la Universidad de Antioquia y del Instituto Colombiano
para la Evaluación de la Educación. El dashboard ADHE es de acceso libre y se
puede consultar en https://fhernanb.shinyapps.io/AppPermanencia/. Los principales
resultados obtenidos fueron que el estrato socioeconómico, el género y el tipo de colegio
están asociados con el fenómeno de deserción. Se encontró que el estrato social 1, los
estudiantes masculinos y el tipo de colegio público tienen las proporciones de deserción
más altas.
El dashboard ADHE podría ser usado por estudiantes,
padres de familia, profesores y administradores para
entender la dinámica de la deserción estudiantil.
64
* Corresponding author: Carmen Elena Patiño-Rodríguez
E-mail: elena.patino@udea.edu.co
ISSN 0120-6230
e-ISSN 2422-2844
DOI: 10.17533/udea.redin.20230519
64
O. D Rivera-Baena et al., Revista Facultad de Ingeniería, Universidad de Antioquia, No. 111, pp. 64-75, 2024
1. Introduction
The methods used to integrate, examine, model, visualize,
and interpret big data are the subjects of what is
known as analytics. Data analytics in higher education
institutions (HEI) provides important prospects to examine,
understand, and model pedagogical processes. However,
one of the main challenges in academic analytics has
been integrating large data volumes that come in diverse
formats from different academic sources and often need to
communicate with each other. The use of data can improve
higher education (HE) practice by enabling more effective
decision-making based on evidence and formulating
responses to address global trends.
An analytic study can be classified depending on the type
of information that it intends to extract from the data.
For example, the descriptive analysis aims to define a
current situation by depicting and summarizing historical
data on students, teaching, research, policies, and other
administrative processes [1].
The main challenges in integrating big data and
academic analytics are the generation and collection
of data; the integration, transformation, and processing
of data, considering challenges of volume, variety,
variability, velocity, and veracity, among others; and
the construction of analytics tools for visualization to
support decision-making, assess scenarios, measure
performance, and communicate the most likely scenarios
to different HE actors [2].
Due to the complexity and variety of sources of data that
HEI may have, some authors proposed four components
(institutional analytics, information technology (IT)
analytics, academic analytics, and learning analytics) to
develop a conceptual framework that describes big data
in HEI [3]. IT focuses on strategic decision-making, using
policy, instructional, and structural analytics to increase
the ability to make appropriate decisions based on data.
Academic analytics aims to effectively measure, collect,
interpret, report, and share data on operational activities
related to educational programming and identify students’
strengths and weaknesses, whereas learning analytics
centers on the learning process.
The potential of education analytics is very significant.
Using it as a basis for decision-making can be fruitful
as using historical data and information helps not only
understand what occurred but also predict what is most
likely to occur in the future and what preparations are
needed to address those most likely scenarios [4, 5]. It
is necessary to understand the information housed even
internally in HEI. Thus, there are still better research
opportunities in integrating what is known as big data
and academic analytics. There are challenges inherent
in these two areas that have not been addressed in
an integrated way through a structured methodology
that allows transforming large volumes of data into
useful, accessible, and transparent information for
decision-making in HE. Regarding the implementation
of analytic tools in HEI, many of the above studies have
focused on developing data-based tools that are assumed
to be available in most HEI.
Student dropout in higher education (HE) is a complex
phenomenon that has been studied from different
perspectives [6–12]. With analytic tools, it is possible
to measure, collect, interpret, report, and share data
to identify factors that affect the student dropout
phenomenon. Despite the positive outcomes that
analytic tools may produce in identifying student dropout
factors, only some academic program administrators have
adopted these tools.
For this reason, this paper presents the general design
and implementation of an academic analytics dashboard in
HE, called ADHE, to support the decision-making process
of educational program administrators.
In Colombia, a public institution is responsible for
evaluating the country’s education quality (Colombian
Institute for the Evaluation of Education - ICFES). This
evaluation is assessed through national tests administered
to students at all educational levels in the country. ICFES
conducts tests in the third and fifth years of primary
education called Saber 3° and Saber 5°, a test in the
fourth year of secondary education called Saber 9°, and a
test in the last year of secondary education called Saber
11. This test assesses areas of mathematics, language,
social sciences, natural sciences, and English language.
Notably, during the tests, sociodemographic and economic
data are collected; these data supplement any analysis
resulting from the evaluation of academic performance.
The results of all these HE quality assessment tests are
stored in ICFES databases. Universidad de Antioquia,
where the analysis presented in this document takes
place, has specific admission tests or criteria. In this case,
the specific admission test consists of two competencies:
reading comprehension and logical reasoning. The
results of admission tests are stored in UDEA databases.
Thus, the potential for integrating information from
different sources, not necessarily within the same HE
institution, is evident. Then we integrated information and
development ADHE, an Academic Analytic Dashboard in
Higher Education.
This paper is divided as follows. The first section
presents theoretical fundamentals to analyze HE data; in
this section, we describe the background of the research
on analytics in HE and related work on academic analytics
and dashboards in HE. The second section presented the
65
1. Introduction
The methods used to integrate, examine, model, visualize,
and interpret big data are the subjects of what is
known as analytics. Data analytics in higher education
institutions (HEI) provides important prospects to examine,
understand, and model pedagogical processes. However,
one of the main challenges in academic analytics has
been integrating large data volumes that come in diverse
formats from different academic sources and often need to
communicate with each other. The use of data can improve
higher education (HE) practice by enabling more effective
decision-making based on evidence and formulating
responses to address global trends.
An analytic study can be classified depending on the type
of information that it intends to extract from the data.
For example, the descriptive analysis aims to define a
current situation by depicting and summarizing historical
data on students, teaching, research, policies, and other
administrative processes [1].
The main challenges in integrating big data and
academic analytics are the generation and collection
of data; the integration, transformation, and processing
of data, considering challenges of volume, variety,
variability, velocity, and veracity, among others; and
the construction of analytics tools for visualization to
support decision-making, assess scenarios, measure
performance, and communicate the most likely scenarios
to different HE actors [2].
Due to the complexity and variety of sources of data that
HEI may have, some authors proposed four components
(institutional analytics, information technology (IT)
analytics, academic analytics, and learning analytics) to
develop a conceptual framework that describes big data
in HEI [3]. IT focuses on strategic decision-making, using
policy, instructional, and structural analytics to increase
the ability to make appropriate decisions based on data.
Academic analytics aims to effectively measure, collect,
interpret, report, and share data on operational activities
related to educational programming and identify students’
strengths and weaknesses, whereas learning analytics
centers on the learning process.
The potential of education analytics is very significant.
Using it as a basis for decision-making can be fruitful
as using historical data and information helps not only
understand what occurred but also predict what is most
likely to occur in the future and what preparations are
needed to address those most likely scenarios [4, 5]. It
is necessary to understand the information housed even
internally in HEI. Thus, there are still better research
opportunities in integrating what is known as big data
and academic analytics. There are challenges inherent
in these two areas that have not been addressed in
an integrated way through a structured methodology
that allows transforming large volumes of data into
useful, accessible, and transparent information for
decision-making in HE. Regarding the implementation
of analytic tools in HEI, many of the above studies have
focused on developing data-based tools that are assumed
to be available in most HEI.
Student dropout in higher education (HE) is a complex
phenomenon that has been studied from different
perspectives [6–12]. With analytic tools, it is possible
to measure, collect, interpret, report, and share data
to identify factors that affect the student dropout
phenomenon. Despite the positive outcomes that
analytic tools may produce in identifying student dropout
factors, only some academic program administrators have
adopted these tools.
For this reason, this paper presents the general design
and implementation of an academic analytics dashboard in
HE, called ADHE, to support the decision-making process
of educational program administrators.
In Colombia, a public institution is responsible for
evaluating the country’s education quality (Colombian
Institute for the Evaluation of Education - ICFES). This
evaluation is assessed through national tests administered
to students at all educational levels in the country. ICFES
conducts tests in the third and fifth years of primary
education called Saber 3° and Saber 5°, a test in the
fourth year of secondary education called Saber 9°, and a
test in the last year of secondary education called Saber
11. This test assesses areas of mathematics, language,
social sciences, natural sciences, and English language.
Notably, during the tests, sociodemographic and economic
data are collected; these data supplement any analysis
resulting from the evaluation of academic performance.
The results of all these HE quality assessment tests are
stored in ICFES databases. Universidad de Antioquia,
where the analysis presented in this document takes
place, has specific admission tests or criteria. In this case,
the specific admission test consists of two competencies:
reading comprehension and logical reasoning. The
results of admission tests are stored in UDEA databases.
Thus, the potential for integrating information from
different sources, not necessarily within the same HE
institution, is evident. Then we integrated information and
development ADHE, an Academic Analytic Dashboard in
Higher Education.
This paper is divided as follows. The first section
presents theoretical fundamentals to analyze HE data; in
this section, we describe the background of the research
on analytics in HE and related work on academic analytics
and dashboards in HE. The second section presented the
65
O. D Rivera-Baena et al., Revista Facultad de Ingeniería, Universidad de Antioquia, No. 111, pp. 64-75, 2024
methodological framework, including information such as
the big data processing pipeline and dashboard planning.
In the third section, we present the construction of an
information visualization tool as a Dashboard and the
main findings related to the student dropout problem.
The fourth section corresponds to discussion results
about three topics: 1) the impact of the academic analytic
Dashboard in HE, 2) the integration of information in HE
from different sources, and 3) the characterization of
student dropout in an HEI. Finally, we presented the main
conclusions and suggestions for future work.
2. Theoretical fundaments to HE data
Several features characterize data as big data in HE.
Some authors discuss certain key features, including 1) a
large amount of information about academic and learning
processes and socioeconomic and academic student
characteristics through longitudinal student data [4].
The information must be stored, processed, transferred,
analyzed, and presented, for example, to examine student
performance patterns over time. Also, 2) HEI data can be
updated and generated frequently due to admission and
assessment processes, graduation, dropout, etc. Finally,
3) data are in diverse structured and unstructured formats
that are generated in teaching, learning, and assessment
activities. These characteristics make HE an area where
analytics can be beneficial for exploiting and classifying
complex information found in large and diverse data sets.
2.1 Analytics in HE
Analytics applied to education have had various objectives:
1) To use analytics to facilitate the initial processing of
data through the integration of information sources and
technological subsystems [5]; 2) to predict the academic
performance of students according to their context
(social, family, economic, etc.) [13] ; 3) to analyze the
degree of association of the variables that influence
students’ performances [14]; and 4) to develop interactive
information visualization tools that, taking as their scope
the development of exploratory analyses of academic
variables, serves as an input for decision-making that can
increase efficiency in HEI [15].
Many of the studies in educational analytics have been
concerned with collecting and analyzing information on
student academic performance, student effort, and the
demographic context of each student. Other studies have
developed student dropout analysis management systems
in Engineering programs, helping to determine student
dropout factors methodologically [16, 17]. However,
regarding information from preuniversity academic
performance, social behaviors, and possible feedback
from teachers and instructors, few studies can be found
[11, 14–16, 18–20].
Some authors have studied the aggregation of information
from the different technological subsystems of a university,
identifying the potential of education analytics for
use in decision-making and improving management
activities related to student performance and institutional
and administrative issues; furthermore, synthesized
successful analytics practices in HE in different institutions
and highlighted relevant aspects, such as the fact that
sources and types of data used in education analytics
have changed dramatically over the years; the success
of analytical studies in this field may depend on the
effectiveness of the integration of data, which not only
come from different sources and are structured in different
ways but are also generated in large volumes. The more
data there are and the more diverse they are, the better
and more fruitful the results will be. Other authors noted
the importance of integrating and visualizing information
through an analytical tool and suggested a feedback
process to allow automated warnings so HEI can make
timely and effective responses [5, 9, 10].
2.2 Types of analytics in HE
HEI should be able to make analytics actionable,
implementable, and executable [21]. Performing
descriptive analytics with these characteristics requires
the capacity to collect data, measure performance
and monitor performance constantly to obtain an
evaluative overview of programs and the institution.
The context of predictive analytics requires projecting
and analyzing relationships between variables and events
to gain a comprehensive understanding of information.
This involves translating the data into correlation and
regression models, which can then be integrated into the
decision-making process. Finally, in prescriptive analytics,
it is essential to create and use optimization and decision
models to guide the implementation of alternatives with
a more significant impact on the objectives and to discard
those with less impact. Understanding the complexity
associated with analytics tools, from data collection to
constructing these tools to obtain useful results, is the
most important step in integrating analytics into strategy
[22, 23].
The tools of descriptive analytics have great potential
that must still be explored. As evidenced in studies,
graphical tools designed to display exploratory visual
analyses not only help test hypotheses more easily
but also support decision-making and ensure proper
monitoring of information [15]. Furthermore, the analytics
tools applied in education are not limited to a single
objective. These studies include tools for data processing
that address the challenges associated with big data
66
methodological framework, including information such as
the big data processing pipeline and dashboard planning.
In the third section, we present the construction of an
information visualization tool as a Dashboard and the
main findings related to the student dropout problem.
The fourth section corresponds to discussion results
about three topics: 1) the impact of the academic analytic
Dashboard in HE, 2) the integration of information in HE
from different sources, and 3) the characterization of
student dropout in an HEI. Finally, we presented the main
conclusions and suggestions for future work.
2. Theoretical fundaments to HE data
Several features characterize data as big data in HE.
Some authors discuss certain key features, including 1) a
large amount of information about academic and learning
processes and socioeconomic and academic student
characteristics through longitudinal student data [4].
The information must be stored, processed, transferred,
analyzed, and presented, for example, to examine student
performance patterns over time. Also, 2) HEI data can be
updated and generated frequently due to admission and
assessment processes, graduation, dropout, etc. Finally,
3) data are in diverse structured and unstructured formats
that are generated in teaching, learning, and assessment
activities. These characteristics make HE an area where
analytics can be beneficial for exploiting and classifying
complex information found in large and diverse data sets.
2.1 Analytics in HE
Analytics applied to education have had various objectives:
1) To use analytics to facilitate the initial processing of
data through the integration of information sources and
technological subsystems [5]; 2) to predict the academic
performance of students according to their context
(social, family, economic, etc.) [13] ; 3) to analyze the
degree of association of the variables that influence
students’ performances [14]; and 4) to develop interactive
information visualization tools that, taking as their scope
the development of exploratory analyses of academic
variables, serves as an input for decision-making that can
increase efficiency in HEI [15].
Many of the studies in educational analytics have been
concerned with collecting and analyzing information on
student academic performance, student effort, and the
demographic context of each student. Other studies have
developed student dropout analysis management systems
in Engineering programs, helping to determine student
dropout factors methodologically [16, 17]. However,
regarding information from preuniversity academic
performance, social behaviors, and possible feedback
from teachers and instructors, few studies can be found
[11, 14–16, 18–20].
Some authors have studied the aggregation of information
from the different technological subsystems of a university,
identifying the potential of education analytics for
use in decision-making and improving management
activities related to student performance and institutional
and administrative issues; furthermore, synthesized
successful analytics practices in HE in different institutions
and highlighted relevant aspects, such as the fact that
sources and types of data used in education analytics
have changed dramatically over the years; the success
of analytical studies in this field may depend on the
effectiveness of the integration of data, which not only
come from different sources and are structured in different
ways but are also generated in large volumes. The more
data there are and the more diverse they are, the better
and more fruitful the results will be. Other authors noted
the importance of integrating and visualizing information
through an analytical tool and suggested a feedback
process to allow automated warnings so HEI can make
timely and effective responses [5, 9, 10].
2.2 Types of analytics in HE
HEI should be able to make analytics actionable,
implementable, and executable [21]. Performing
descriptive analytics with these characteristics requires
the capacity to collect data, measure performance
and monitor performance constantly to obtain an
evaluative overview of programs and the institution.
The context of predictive analytics requires projecting
and analyzing relationships between variables and events
to gain a comprehensive understanding of information.
This involves translating the data into correlation and
regression models, which can then be integrated into the
decision-making process. Finally, in prescriptive analytics,
it is essential to create and use optimization and decision
models to guide the implementation of alternatives with
a more significant impact on the objectives and to discard
those with less impact. Understanding the complexity
associated with analytics tools, from data collection to
constructing these tools to obtain useful results, is the
most important step in integrating analytics into strategy
[22, 23].
The tools of descriptive analytics have great potential
that must still be explored. As evidenced in studies,
graphical tools designed to display exploratory visual
analyses not only help test hypotheses more easily
but also support decision-making and ensure proper
monitoring of information [15]. Furthermore, the analytics
tools applied in education are not limited to a single
objective. These studies include tools for data processing
that address the challenges associated with big data
66
O. D Rivera-Baena et al., Revista Facultad de Ingeniería, Universidad de Antioquia, No. 111, pp. 64-75, 2024
and large volumes of data, facilitating the integration
and manipulation of information from different sources
[5, 14, 15] tools focused on visualization through web
platforms and applications specializing in the visual
descriptive analysis [15]; and machine learning tools,
predictive analytics, probabilistic and supervised learning
models, such as logistic regression, decision trees, and
support vector machines [24].
2.3 Academic analytics in HE
Regarding the specific objectives with which academic
analytics studies have been developed, some prevail
within the academic analytical literature are 1) to facilitate
learning and academic progress; 2) to strengthen the
effectiveness of learning support strategies; and, to a
lesser extent, 3) to improve administrative effectiveness
[12]. Thus, from the analytic perspective, the first objective
seeks to improve academic performance; the second
objective proposes to support students in the early stages
of their programs to ensure academic success and provide
information to generate warnings to identify and assist
at-risk students; and the third objective attempts to
provide information to improve curriculum design and for
well-informed management decision-making regarding
the recruitment, admission, and retention of students.
The student dropout phenomenon is articulated with
the objectives of academic analytics. Studies focused
on early detection of high-risk status among students
who are generally at higher risk of dropping out of HE
has been developed [10, 19–23]. This enabled enhancing
early strategies to support and intervene with the most
vulnerable students, increasing their academic success
and the effectiveness of the strategies themselves.
Other authors have studied strategies for supporting
permanence and intervention. Understanding the factors
that affect student retention has become essential in
the retention analysis. The study analyzes academic
and psychological factors through structural equation
modeling.
In business intelligence, the clean-up and preprocessing
stage of data is crucial in the analytics processes, and it is
not accidental that it is one of the most extensive stages.
It also assesses the potential of dashboards, constructed
and processed appropriately, to serve as a pivotal tool and
offer dynamic yet impartial environments that cater to
the informational requirements of decision-makers and
end-users alike [28].
2.4 Dashboards in HE
Studies that have developed dashboards in HE have
focused on learning analytics [29, 30] and academic
analytics [15, 22, 31, 32]. In learning analytics, VisMOOC
[29] is a visual analytic system to help analyze user learning
behaviors by using video clickstream data from MOOC
platforms and exploring video utilization from multiple
perspectives. In academic analytics, the visual analytics
tool for exploratory analysis of Academic Analytics.
The tool supports various interactive data visualization
methods and develops a web platform capable of managing
the metadata of medical and health programs, with
constant updates that support curricular innovations and
their adoption within academic programs to increase their
effectiveness. Additionally, visual analytics systems not
only help instructors and education experts understand the
reasons for student dropout but also allow researchers
to identify crucial features that can further improve the
performance of the models. Moreover, although it is a
much more theoretical tool, it serves as a starting point
to apply analytics tools and highlights the need to propose
generalizable strategies for many more processes in HE.
3. Methodology
To understand the relationship between student dropout
and academic and socioeconomic factors of students
in HEI, we used the methodology of named “big data
processing pipeline” [2], slightly modified for this study.
The methodology shown in Figure 1, considers five
phases: data acquisition, data integration, data cleaning,
data transformation, and finally, visualization.
Figure 1 Big data processing pipeline. Source: Own elaboration
based on [2]
The data analyzed in this paper were taken from the data
repository of the Colombian Institute for the Evaluation of
Education, DataIcfes [33], and the institutional information
system of Universidad de Antioquia. The data analyzed
comes from three databases. The first set of databases
was obtained from the DataIcfes repository, and it contains
the results of the Saber 11 test between 20061 and 20172
and factors associated with the student and school. The
second database, the admission dataset, was obtained
from the institutional information system of Universidad de
Antioquia and includes the results of the entrance tests of
applicants to programs of the engineering faculty between
2010 and 2018, in addition to factors associated with the
students and with admission. The third database, referred
to as the dropout dataset, comprises data about the
academic performance of students within the engineering
faculty from 2010 to 2018, along with factors related to
67
and large volumes of data, facilitating the integration
and manipulation of information from different sources
[5, 14, 15] tools focused on visualization through web
platforms and applications specializing in the visual
descriptive analysis [15]; and machine learning tools,
predictive analytics, probabilistic and supervised learning
models, such as logistic regression, decision trees, and
support vector machines [24].
2.3 Academic analytics in HE
Regarding the specific objectives with which academic
analytics studies have been developed, some prevail
within the academic analytical literature are 1) to facilitate
learning and academic progress; 2) to strengthen the
effectiveness of learning support strategies; and, to a
lesser extent, 3) to improve administrative effectiveness
[12]. Thus, from the analytic perspective, the first objective
seeks to improve academic performance; the second
objective proposes to support students in the early stages
of their programs to ensure academic success and provide
information to generate warnings to identify and assist
at-risk students; and the third objective attempts to
provide information to improve curriculum design and for
well-informed management decision-making regarding
the recruitment, admission, and retention of students.
The student dropout phenomenon is articulated with
the objectives of academic analytics. Studies focused
on early detection of high-risk status among students
who are generally at higher risk of dropping out of HE
has been developed [10, 19–23]. This enabled enhancing
early strategies to support and intervene with the most
vulnerable students, increasing their academic success
and the effectiveness of the strategies themselves.
Other authors have studied strategies for supporting
permanence and intervention. Understanding the factors
that affect student retention has become essential in
the retention analysis. The study analyzes academic
and psychological factors through structural equation
modeling.
In business intelligence, the clean-up and preprocessing
stage of data is crucial in the analytics processes, and it is
not accidental that it is one of the most extensive stages.
It also assesses the potential of dashboards, constructed
and processed appropriately, to serve as a pivotal tool and
offer dynamic yet impartial environments that cater to
the informational requirements of decision-makers and
end-users alike [28].
2.4 Dashboards in HE
Studies that have developed dashboards in HE have
focused on learning analytics [29, 30] and academic
analytics [15, 22, 31, 32]. In learning analytics, VisMOOC
[29] is a visual analytic system to help analyze user learning
behaviors by using video clickstream data from MOOC
platforms and exploring video utilization from multiple
perspectives. In academic analytics, the visual analytics
tool for exploratory analysis of Academic Analytics.
The tool supports various interactive data visualization
methods and develops a web platform capable of managing
the metadata of medical and health programs, with
constant updates that support curricular innovations and
their adoption within academic programs to increase their
effectiveness. Additionally, visual analytics systems not
only help instructors and education experts understand the
reasons for student dropout but also allow researchers
to identify crucial features that can further improve the
performance of the models. Moreover, although it is a
much more theoretical tool, it serves as a starting point
to apply analytics tools and highlights the need to propose
generalizable strategies for many more processes in HE.
3. Methodology
To understand the relationship between student dropout
and academic and socioeconomic factors of students
in HEI, we used the methodology of named “big data
processing pipeline” [2], slightly modified for this study.
The methodology shown in Figure 1, considers five
phases: data acquisition, data integration, data cleaning,
data transformation, and finally, visualization.
Figure 1 Big data processing pipeline. Source: Own elaboration
based on [2]
The data analyzed in this paper were taken from the data
repository of the Colombian Institute for the Evaluation of
Education, DataIcfes [33], and the institutional information
system of Universidad de Antioquia. The data analyzed
comes from three databases. The first set of databases
was obtained from the DataIcfes repository, and it contains
the results of the Saber 11 test between 20061 and 20172
and factors associated with the student and school. The
second database, the admission dataset, was obtained
from the institutional information system of Universidad de
Antioquia and includes the results of the entrance tests of
applicants to programs of the engineering faculty between
2010 and 2018, in addition to factors associated with the
students and with admission. The third database, referred
to as the dropout dataset, comprises data about the
academic performance of students within the engineering
faculty from 2010 to 2018, along with factors related to
67
O. D Rivera-Baena et al., Revista Facultad de Ingeniería, Universidad de Antioquia, No. 111, pp. 64-75, 2024
Table 1 Number of students in each database
Database Number of students
DataIcfes 6.579.352
Admission 30.548
Dropout 6.859
Table 2 Variables in database
Type Variable
Demographic Gender
Family background Father’s educational level
Mother’s educational level
Financial Family income
Social stratum
Pre-enrollment Saber 11 test scores
School emphasis
School type
Monthly tuition payment
Social stratification in school
Multidimensional poverty
index of high school
Enrollment Reading comprehension score
Logic reasoning score
Admission year
Admission type
Admission program
Semester-related Semester GPA
Range
Academic level
Other programs
both the students themselves and the program in which
they were enrolled. Table 1 shows the number of students
in each database.
In the data integration phase, the three databases were
related. In the first instance, the dropout and admission
databases were related; the student information, the
engineering program to which they were admitted, and
the school code were used as a filter. Once these two
databases were related, we proceeded to relate them
to the Saber 11 database and used student information
and school code as a filter. In the data transformation
phase, some variables were transformed and created.
The number of students in this database was 6.593,
registered in 12 on-campus programs, 11 distance
programs, and four online programs. Table 2 shows the
dataset variables after the fourth phase of the big data
processing pipeline. The database contains demographic,
family background, financial, pre-enrollment, enrollment,
and semester-related variables. Figure 2 presents the
methodology used to develop the dashboard to visualize
the information obtained from the final dropout database.
The proposed methodology consists of a series of steps
that include consolidation of the database, planning of
the visualization structure, prototype development, and
dashboard improvement using feedback from users and
decision-makers.
Within the planning of the visualization structure, the
following methodology was carried out for the final
graphical display of the results of the descriptive analytics
study. First, the different hypotheses that the data could
help answer were posited, and based on this, other
sections were conceived for the dashboard. There would
be a section that grouped the different academic variables
for the whole faculty, for each program, and for each
academic level; in this way, academic performance could
be related to the student dropout problem. Another section
will present the results of the knowledge assessment
tests that everyone must pass before having the status
of a university student and subsequently, the status of
having dropped out of HE. A third section would relate the
students’ social, economic, and family variables to give
them context and understand the different dimensions of
the student dropout phenomenon in HE.
The dashboard was developed in the Shiny package
of the programming language and environment for
statistical computing R. To construct the dashboard, we
used functions of the Shiny [34], shinydashboard [35],
shinyWidgets [36], ggplot2 [37], plotly [38], dplyr [39], DT
[40], and viridis [41] packages. Finally, the feedback was
carried out with the Department Directors, the Curriculum
Development Committee, and the Vice Dean of the Faculty
under study, where the importance of having a variable
that would allow segmenting and analyzing the results
according to specific periods of interest was evidenced.
4. Results
The construction of an information visualization tool is
the main technical result of the proposed methodology.
The visualization tool ADPHE allows for transforming
large volumes of data from different sources into
visual information relevant to decision-making and
understanding the student dropout phenomenon in HE.
Thanks to the interactive and graphical structure of the
dashboard, it is feasible to effectively communicate
the most pertinent aspects of the problem to all
stakeholders of the academic community, including
students, parents, teachers, administrators, and so
on. The ADPHE dashboard was built using the R
language programming R, and it is hosted at shinyapp.io
service, which is available to any user at the URL
https://fhernanb.shinyapps.io/AppPermanencia/. The
structure of the developed ADPHE dashboard is shown in
Figure 3.
Marker A in Figure 3 groups together the four main
sections of the application: Summary, Dropout,
Vulnerability, and Exams and Vulnerability. In the
Summary section, proportional and absolute frequency
68
Table 1 Number of students in each database
Database Number of students
DataIcfes 6.579.352
Admission 30.548
Dropout 6.859
Table 2 Variables in database
Type Variable
Demographic Gender
Family background Father’s educational level
Mother’s educational level
Financial Family income
Social stratum
Pre-enrollment Saber 11 test scores
School emphasis
School type
Monthly tuition payment
Social stratification in school
Multidimensional poverty
index of high school
Enrollment Reading comprehension score
Logic reasoning score
Admission year
Admission type
Admission program
Semester-related Semester GPA
Range
Academic level
Other programs
both the students themselves and the program in which
they were enrolled. Table 1 shows the number of students
in each database.
In the data integration phase, the three databases were
related. In the first instance, the dropout and admission
databases were related; the student information, the
engineering program to which they were admitted, and
the school code were used as a filter. Once these two
databases were related, we proceeded to relate them
to the Saber 11 database and used student information
and school code as a filter. In the data transformation
phase, some variables were transformed and created.
The number of students in this database was 6.593,
registered in 12 on-campus programs, 11 distance
programs, and four online programs. Table 2 shows the
dataset variables after the fourth phase of the big data
processing pipeline. The database contains demographic,
family background, financial, pre-enrollment, enrollment,
and semester-related variables. Figure 2 presents the
methodology used to develop the dashboard to visualize
the information obtained from the final dropout database.
The proposed methodology consists of a series of steps
that include consolidation of the database, planning of
the visualization structure, prototype development, and
dashboard improvement using feedback from users and
decision-makers.
Within the planning of the visualization structure, the
following methodology was carried out for the final
graphical display of the results of the descriptive analytics
study. First, the different hypotheses that the data could
help answer were posited, and based on this, other
sections were conceived for the dashboard. There would
be a section that grouped the different academic variables
for the whole faculty, for each program, and for each
academic level; in this way, academic performance could
be related to the student dropout problem. Another section
will present the results of the knowledge assessment
tests that everyone must pass before having the status
of a university student and subsequently, the status of
having dropped out of HE. A third section would relate the
students’ social, economic, and family variables to give
them context and understand the different dimensions of
the student dropout phenomenon in HE.
The dashboard was developed in the Shiny package
of the programming language and environment for
statistical computing R. To construct the dashboard, we
used functions of the Shiny [34], shinydashboard [35],
shinyWidgets [36], ggplot2 [37], plotly [38], dplyr [39], DT
[40], and viridis [41] packages. Finally, the feedback was
carried out with the Department Directors, the Curriculum
Development Committee, and the Vice Dean of the Faculty
under study, where the importance of having a variable
that would allow segmenting and analyzing the results
according to specific periods of interest was evidenced.
4. Results
The construction of an information visualization tool is
the main technical result of the proposed methodology.
The visualization tool ADPHE allows for transforming
large volumes of data from different sources into
visual information relevant to decision-making and
understanding the student dropout phenomenon in HE.
Thanks to the interactive and graphical structure of the
dashboard, it is feasible to effectively communicate
the most pertinent aspects of the problem to all
stakeholders of the academic community, including
students, parents, teachers, administrators, and so
on. The ADPHE dashboard was built using the R
language programming R, and it is hosted at shinyapp.io
service, which is available to any user at the URL
https://fhernanb.shinyapps.io/AppPermanencia/. The
structure of the developed ADPHE dashboard is shown in
Figure 3.
Marker A in Figure 3 groups together the four main
sections of the application: Summary, Dropout,
Vulnerability, and Exams and Vulnerability. In the
Summary section, proportional and absolute frequency
68
O. D Rivera-Baena et al., Revista Facultad de Ingeniería, Universidad de Antioquia, No. 111, pp. 64-75, 2024
Figure 2 Methodology for building a dashboard
Figure 3 Structure of the ADPHE dashboard
analyses are presented on some socioeconomic variables
generally associated with the student dropout problem;
in this section, the most important information is
summarized through bar charts, bubble charts, and
frequency tables. The Dropout section is divided into
three parts. “Faculty” allows evaluating the relationships
between up to two academic and socioeconomic variables,
and “Academic Program” enables analyzing different
academic variables for each on-campus academic
program of the engineering faculty, using graphs and
key performance indicators, such as dropout proportion
and dropout average GPA. Finally, “Year” presents a
history of the dropout proportion by academic year and
program from the 2011-1 academic semester to the
2017-2 semester. It also allows comparing up to two
academic programs for different time windows. The
Exams and Vulnerability section provides information on
the scores obtained by students on each of the admission
tests and the Saber 11 test, according to their family
and socioeconomic context. In this way, it is possible to
determine whether the conditions of the individual who
dropped out of their HE program could have ultimately
conditioned their performance on the tests. Finally, the
Vulnerability section allows comparing two important
variables: the type of admission and academic level
of dropout students, with many of the variables that
characterize students socially and economically.
In contrast, marker B in Figure 3 shows the two
complementary sections of the dashboard. The About
the Project section presents the study’s technical file and
relevant information for researchers and decision-makers.
The Glossary section contains all the important terms to
understand the graphics and information presented in the
dashboard.
An example of how the drop-down lists of variables
to analyze are presented in each of the sections is
presented in Figure 3, specifically in the C marker, where
in this case, the social stratum variable is selected as
an example. Marker D exemplifies another drop-down
list type in the application; these lists are not related to
variables but rather to how the users want to visualize
the information. Finally, the E marker shows that each
section contains a tab box with supporting information or
with display alternatives; in this case, the information can
be visualized using frequency tables or alternating to a
bubble chart.
69
Figure 2 Methodology for building a dashboard
Figure 3 Structure of the ADPHE dashboard
analyses are presented on some socioeconomic variables
generally associated with the student dropout problem;
in this section, the most important information is
summarized through bar charts, bubble charts, and
frequency tables. The Dropout section is divided into
three parts. “Faculty” allows evaluating the relationships
between up to two academic and socioeconomic variables,
and “Academic Program” enables analyzing different
academic variables for each on-campus academic
program of the engineering faculty, using graphs and
key performance indicators, such as dropout proportion
and dropout average GPA. Finally, “Year” presents a
history of the dropout proportion by academic year and
program from the 2011-1 academic semester to the
2017-2 semester. It also allows comparing up to two
academic programs for different time windows. The
Exams and Vulnerability section provides information on
the scores obtained by students on each of the admission
tests and the Saber 11 test, according to their family
and socioeconomic context. In this way, it is possible to
determine whether the conditions of the individual who
dropped out of their HE program could have ultimately
conditioned their performance on the tests. Finally, the
Vulnerability section allows comparing two important
variables: the type of admission and academic level
of dropout students, with many of the variables that
characterize students socially and economically.
In contrast, marker B in Figure 3 shows the two
complementary sections of the dashboard. The About
the Project section presents the study’s technical file and
relevant information for researchers and decision-makers.
The Glossary section contains all the important terms to
understand the graphics and information presented in the
dashboard.
An example of how the drop-down lists of variables
to analyze are presented in each of the sections is
presented in Figure 3, specifically in the C marker, where
in this case, the social stratum variable is selected as
an example. Marker D exemplifies another drop-down
list type in the application; these lists are not related to
variables but rather to how the users want to visualize
the information. Finally, the E marker shows that each
section contains a tab box with supporting information or
with display alternatives; in this case, the information can
be visualized using frequency tables or alternating to a
bubble chart.
69
O. D Rivera-Baena et al., Revista Facultad de Ingeniería, Universidad de Antioquia, No. 111, pp. 64-75, 2024
4.1 Main findings related to the student
dropout problem
It is commonly hypothesized that economic conditions
determine whether a student will drop out of HE over
time. It seems almost a preconceived idea that the
higher the student’s socioeconomic stratum (better their
socioeconomic conditions), the lower their probability of
dropping out. As shown in Figure 4, one of the main results
in the dashboard is that, when a proportional analysis of
the dropouts is carried out, the ratios of student dropouts
do differ greatly among socioeconomic strata. We applied
a Pearson’s chi-squared test to compare the dropout
proportions through social stratum and found a p-value
= 1.485 × 10−11, meaning there is a difference between
the proportions. This shows that socioeconomic level is a
determining factor in the student dropout phenomenon,
as is commonly thought. This pattern is similar to other
factors, such as gender (p -value < 2.2 × 10−16) and
type of high school (p -value = 3.653 × 10−9) where
the student studied; in these last variables, there is less
difference between proportions.
Figure 4 Dropout proportion by social stratum
Another important result is presented in Figure 5, which
shows the dropout proportion by type of admission to
the university. Notably, the type of general admission is
through the admission test (POR-EXAM), where the two
measured competencies are logical reasoning and reading
comprehension. However, the figure presents up to 13
other types of admission, most structured as equitable
admission strategies for groups generally considered
vulnerable, such as negritude (NEGRITUD) and indigenous
(INDIGENA) groups. In these types of equitable admission
strategies, the dropout proportion is much higher than the
proportion of the general exam method. This indicates
that although inclusion efforts exist, they do not remain
effective over time, since many of these vulnerable
students also end up dropping out. Types of admission
related to a change of program (CAMB-PRG) or modality
(CAMBMODA) have the lowest dropout proportion since,
generally, when a student decides to leave one academic
program to enter another, it is because he or she has gone
through a process of internal reflection and is firm about
what he or she truly wants for his or her professional
life, and such students rarely leave an academic program
again.
Notably, the type of admission with the highest dropout
proportion is AJUPEI; the university conceived this
admission strategy as an alternative for those who obtain
scores close to the admission exam cut-off score but still
need to be admitted. This strategy is aimed at helping
people in rural areas, and people registered for admission
in seats of the university other than the main campus.
However, this admission strategy has yet to prove effective,
and decision-makers should re-evaluate it or implement
complementary strategies to guarantee the permanence
of those who enter with a lower performance.
Figure 5 Dropout proportion by type of admission
Figure 6 presents the distribution of the cumulative GPAs
of students who dropped out of their academic program
and who finished their educational program, about the
type of university admission. In the case of those who
dropped out of their academic program, it is shown
that for almost all types of admission, at least 25% of
students (Q3) dropped out with a cumulative GPA above
the passing grade (3.0); that is, at least 25% of students
in almost all types of admissions dropped out for reasons
beyond academics, which might include motivational
or curriculum-related factors. In this case, we applied
the Kruskal Wallis test to verify the difference between
the type of admission according to the GPA of students
who dropped out, obtaining as a response that there is a
difference in the GPA according to the type of admission.
In the case of those who finished their academic program,
it is shown that for almost all types of admission, at least
25% of those who completed their academic program had
a cumulative GPA above 3.5. As in the previous case, the
Kruskal Wallis test indicated that there is a difference
between the type of admission according to the GPA of
graduated students. As expected, the figure shows that
the GPA boxplot for graduated students lies above the GPA
boxplot for dropout students.
Regarding the academic dropout level, Figure 7 shows the
panorama of an academic program - in this case, industrial
engineering. However, the pattern is generalizable to
the other academic programs of the engineering faculty.
Moreover, it is a pattern in which early student dropout
70
4.1 Main findings related to the student
dropout problem
It is commonly hypothesized that economic conditions
determine whether a student will drop out of HE over
time. It seems almost a preconceived idea that the
higher the student’s socioeconomic stratum (better their
socioeconomic conditions), the lower their probability of
dropping out. As shown in Figure 4, one of the main results
in the dashboard is that, when a proportional analysis of
the dropouts is carried out, the ratios of student dropouts
do differ greatly among socioeconomic strata. We applied
a Pearson’s chi-squared test to compare the dropout
proportions through social stratum and found a p-value
= 1.485 × 10−11, meaning there is a difference between
the proportions. This shows that socioeconomic level is a
determining factor in the student dropout phenomenon,
as is commonly thought. This pattern is similar to other
factors, such as gender (p -value < 2.2 × 10−16) and
type of high school (p -value = 3.653 × 10−9) where
the student studied; in these last variables, there is less
difference between proportions.
Figure 4 Dropout proportion by social stratum
Another important result is presented in Figure 5, which
shows the dropout proportion by type of admission to
the university. Notably, the type of general admission is
through the admission test (POR-EXAM), where the two
measured competencies are logical reasoning and reading
comprehension. However, the figure presents up to 13
other types of admission, most structured as equitable
admission strategies for groups generally considered
vulnerable, such as negritude (NEGRITUD) and indigenous
(INDIGENA) groups. In these types of equitable admission
strategies, the dropout proportion is much higher than the
proportion of the general exam method. This indicates
that although inclusion efforts exist, they do not remain
effective over time, since many of these vulnerable
students also end up dropping out. Types of admission
related to a change of program (CAMB-PRG) or modality
(CAMBMODA) have the lowest dropout proportion since,
generally, when a student decides to leave one academic
program to enter another, it is because he or she has gone
through a process of internal reflection and is firm about
what he or she truly wants for his or her professional
life, and such students rarely leave an academic program
again.
Notably, the type of admission with the highest dropout
proportion is AJUPEI; the university conceived this
admission strategy as an alternative for those who obtain
scores close to the admission exam cut-off score but still
need to be admitted. This strategy is aimed at helping
people in rural areas, and people registered for admission
in seats of the university other than the main campus.
However, this admission strategy has yet to prove effective,
and decision-makers should re-evaluate it or implement
complementary strategies to guarantee the permanence
of those who enter with a lower performance.
Figure 5 Dropout proportion by type of admission
Figure 6 presents the distribution of the cumulative GPAs
of students who dropped out of their academic program
and who finished their educational program, about the
type of university admission. In the case of those who
dropped out of their academic program, it is shown
that for almost all types of admission, at least 25% of
students (Q3) dropped out with a cumulative GPA above
the passing grade (3.0); that is, at least 25% of students
in almost all types of admissions dropped out for reasons
beyond academics, which might include motivational
or curriculum-related factors. In this case, we applied
the Kruskal Wallis test to verify the difference between
the type of admission according to the GPA of students
who dropped out, obtaining as a response that there is a
difference in the GPA according to the type of admission.
In the case of those who finished their academic program,
it is shown that for almost all types of admission, at least
25% of those who completed their academic program had
a cumulative GPA above 3.5. As in the previous case, the
Kruskal Wallis test indicated that there is a difference
between the type of admission according to the GPA of
graduated students. As expected, the figure shows that
the GPA boxplot for graduated students lies above the GPA
boxplot for dropout students.
Regarding the academic dropout level, Figure 7 shows the
panorama of an academic program - in this case, industrial
engineering. However, the pattern is generalizable to
the other academic programs of the engineering faculty.
Moreover, it is a pattern in which early student dropout
70
O. D Rivera-Baena et al., Revista Facultad de Ingeniería, Universidad de Antioquia, No. 111, pp. 64-75, 2024
Figure 6 Dropout cumulative GPA by type of admission
is predominant; the most significant number of students
drop out of their academic program at levels 1, 2, and 3.
Figure 7 Dropout academic level
Likewise, the indicators presented in Figure 7 show that
in the industrial engineering program, 17.05% of students
dropped out in the first semester, and almost 30% dropped
out in the first half of the academic program (levels 1 to
5). However, only 2.33% of students dropped out once this
threshold was passed.
Analyzing the student dropout phenomenon by academic
year, using the case of the industrial engineering program
again, Figure 8 shows that the dropout proportion is
not constant over time; in some academic semesters,
it is higher, indicating that it is a dynamic phenomenon
that is also affected by temporary problems or by
time-dependent relationships. Moreover, in a specific
semester, it is possible to find social, economic, academic,
curricular, or other changes or phenomena that influence
the increase in the dropout proportion.
In contrast, as shown in Figure 9, the analysis of the
score obtained on the logical reasoning test by students
who dropped out of their academic program concerning
socioeconomic stratum reflects that, among students who
dropped out, those belonging to the highest socioeconomic
stratum scored better than those at the lowest stratum.
Comparing dropouts and graduated students through
the Kruskal Wallis test, we found the logical reasoning
score was different in strata 1, 2, 3, and 4, which indicates
that the performance in this test is a protective factor for
permanence. In the case of social stratum 5, no differences
were observed between the scores obtained by the two
groups, which points out that student dropout in social
stratum 5 is not due to the ability in logical reasoning.
In social stratum 6, despite the high performance in the
logical reasoning test, we observed a student dropout of
100% of the students. Figure 10 shows similar findings
with the mathematics test of the secondary school.
Figure 11 shows a scatter plot between the scores obtained
in the university entrance tests (logical reasoning and
reading comprehension) for dropped out and graduated.
We note that graduated students tend to have equal
scores on both tests, which means that having balanced
scores on the two tests represents a protective factor
for permanence. In contrast, we note that dropped-out
students had unbalanced scores on both tests. This fact
shows that engineering is not only connected to abilities
in quantitative analysis but rather that the two abilities
complement each other.
5. Discussion
This paper presented a methodology to integrate
information from different academic sources and
proposed the ADHE with demographic, academic, social,
and economic information on students before and during
their professional studies at specific higher education
institutions. The objective of the study was to integrate and
visualize student information that would allow academic
administrators to visualize the relationships between
71
Figure 6 Dropout cumulative GPA by type of admission
is predominant; the most significant number of students
drop out of their academic program at levels 1, 2, and 3.
Figure 7 Dropout academic level
Likewise, the indicators presented in Figure 7 show that
in the industrial engineering program, 17.05% of students
dropped out in the first semester, and almost 30% dropped
out in the first half of the academic program (levels 1 to
5). However, only 2.33% of students dropped out once this
threshold was passed.
Analyzing the student dropout phenomenon by academic
year, using the case of the industrial engineering program
again, Figure 8 shows that the dropout proportion is
not constant over time; in some academic semesters,
it is higher, indicating that it is a dynamic phenomenon
that is also affected by temporary problems or by
time-dependent relationships. Moreover, in a specific
semester, it is possible to find social, economic, academic,
curricular, or other changes or phenomena that influence
the increase in the dropout proportion.
In contrast, as shown in Figure 9, the analysis of the
score obtained on the logical reasoning test by students
who dropped out of their academic program concerning
socioeconomic stratum reflects that, among students who
dropped out, those belonging to the highest socioeconomic
stratum scored better than those at the lowest stratum.
Comparing dropouts and graduated students through
the Kruskal Wallis test, we found the logical reasoning
score was different in strata 1, 2, 3, and 4, which indicates
that the performance in this test is a protective factor for
permanence. In the case of social stratum 5, no differences
were observed between the scores obtained by the two
groups, which points out that student dropout in social
stratum 5 is not due to the ability in logical reasoning.
In social stratum 6, despite the high performance in the
logical reasoning test, we observed a student dropout of
100% of the students. Figure 10 shows similar findings
with the mathematics test of the secondary school.
Figure 11 shows a scatter plot between the scores obtained
in the university entrance tests (logical reasoning and
reading comprehension) for dropped out and graduated.
We note that graduated students tend to have equal
scores on both tests, which means that having balanced
scores on the two tests represents a protective factor
for permanence. In contrast, we note that dropped-out
students had unbalanced scores on both tests. This fact
shows that engineering is not only connected to abilities
in quantitative analysis but rather that the two abilities
complement each other.
5. Discussion
This paper presented a methodology to integrate
information from different academic sources and
proposed the ADHE with demographic, academic, social,
and economic information on students before and during
their professional studies at specific higher education
institutions. The objective of the study was to integrate and
visualize student information that would allow academic
administrators to visualize the relationships between
71
O. D Rivera-Baena et al., Revista Facultad de Ingeniería, Universidad de Antioquia, No. 111, pp. 64-75, 2024
Figure 8 Dropout proportion by year
Figure 9 The score obtained on the logical reasoning test vs.
social stratum
Figure 10 The score obtained on the mathematics test vs.
social stratum
student dropout and demographic, academic, and social
factors. The discussion is structured according to three
characteristics: 1) the impact of the academic analytic
dashboard in HE, 2) the integration of information in HE
from different sources, and 3) the characterization of
Figure 11 Scatterplot between the scores in logical reasoning
and reading comprehension. The diagonal line at 45° represents
balanced scores.
dropouts in an HEI.
5.1 Impact of academic analytic
dashboard in HE
This paper shows an ADHE that supports student dropout
analysis. Some of the origins of this research are
supported using open data from ICFES and its integration
into information from the study institution, in addition to the
institution’s interest in understanding the relationships
between student dropout and factors associated with
students. The implementation of ADHE would support
an academic analytic program and, according to [4],
dashboards like this allow identification and appropriate
rectification of operational activities related to academic
programming and student strengths and weaknesses.
Improvements on the ADHE could focus on providing
constant updates of data. Developing tools as
web platforms with constant updates support their
72
Figure 8 Dropout proportion by year
Figure 9 The score obtained on the logical reasoning test vs.
social stratum
Figure 10 The score obtained on the mathematics test vs.
social stratum
student dropout and demographic, academic, and social
factors. The discussion is structured according to three
characteristics: 1) the impact of the academic analytic
dashboard in HE, 2) the integration of information in HE
from different sources, and 3) the characterization of
Figure 11 Scatterplot between the scores in logical reasoning
and reading comprehension. The diagonal line at 45° represents
balanced scores.
dropouts in an HEI.
5.1 Impact of academic analytic
dashboard in HE
This paper shows an ADHE that supports student dropout
analysis. Some of the origins of this research are
supported using open data from ICFES and its integration
into information from the study institution, in addition to the
institution’s interest in understanding the relationships
between student dropout and factors associated with
students. The implementation of ADHE would support
an academic analytic program and, according to [4],
dashboards like this allow identification and appropriate
rectification of operational activities related to academic
programming and student strengths and weaknesses.
Improvements on the ADHE could focus on providing
constant updates of data. Developing tools as
web platforms with constant updates support their
72
O. D Rivera-Baena et al., Revista Facultad de Ingeniería, Universidad de Antioquia, No. 111, pp. 64-75, 2024
adoption within academic programs and increase their
effectiveness. The ADHE could update student information
on a semesterly basis to offer insights into student
attrition analysis, as certain factors may fluctuate over
time. Additionally, the ADHE could possess an in-house
data preprocessing tool to evaluate and enhance the
data quality for dependable analysis. Moreover, data
preprocessing is critical since it enables the extraction of
high-quality output information from large educational
data sources, such as the present case [2, 14].
5.2 Integration of information in HE from
different sources
Incorporating data from various sources should not
be treated as a static process, as data continues to be
generated at an increasingly rapid pace. Information
repositories that simply capture data from a single
moment in time are not effective, and integration must
be ongoing and adaptable to address challenges such
as those presented in [4], which include data velocity
and variety. As a result, it is essential that integration
processes are fast and efficient, and that the dashboards
utilized for presenting information are dynamic over time.
On the other hand, it is crucial for dashboards, as a
tool for descriptive analytics, to serve as a means of
integration for various forms of analytics. Each dashboard
undertakes an internal data processing step to present the
information in an organized, lucid, and practical manner.
However, for a dashboard to be an effective integration
tool, it should facilitate the generation of integrated
data following the internal processing step. Exporting
preprocessed raw data that can serve as input to other
analytics such as predictive and make many strategic
decisions. Unfortunately, ADHE does not include dynamic
integration or data export functions, but it is clear the
importance of them as an opportunity for future work and
research.
5.3 Characterization of student dropout in
an HEI
Dashboards and reports play a fundamental role as tools
for visualization, tracking, and information consolidation.
In this case, the ADHE allows for finding relationships
between dropouts and variables related to demographic,
social, and academic performance. From a marginal point
of view, ADHE showed there is a difference between the
dropout proportion according to social stratum, gender,
and type of high school. Also, the main dropout proportion
occurred before reaching the middle of the program.
This finding allows designing policies and projects to
analyze the phenomenon in the first semesters. Figure
6 indicates that student dropout is not solely caused by
academic factors. The figure displays that a percentage
of students who left the faculty had met the academic
requirements, having attained a cumulative GPA above
the passing grade of 3.0. Consequently, we propose that
HEI could incorporate additional factors such as stress
levels, student poverty, lack of parental involvement, and
lack of self-motivation, as these factors could heighten
the likelihood of student attrition in HE. Moreover, future
research could focus on developing predictive models for
forecasting early student dropout as well as predicting
student attrition on a semester-by-semester basis.
6. Conclusions
In this paper, we propose an academic analytic dashboard
named ADHE, which is a novel tool in academic analytics
because it allows us to integrate data from different
academic sources and depict the relations between
multiple variables interactively. ADHE can be used for
decision-makers to evaluate the relationship between
some variables and student dropout in HEI, and this new
knowledge can be used to set future policies to minimize
student dropout proportion. This visualization tool, the
result of using descriptive analytics, can also serve as a
basis for further studies on predictive and prescriptive
academic analytics. This study contributes by encouraging
HEI to build their dashboards to share useful information
with students, teachers, and administrators.
The integration of information from different sources
requires developing a series of technical and business
intelligence processes in a structured manner, given
that only through an established methodology is it
possible to convert large volumes of data - which are also
variable, heterogeneous, and generated at high velocity
- into reliable, and valuable information. An accurate
integration process facilitates the comprehension,
cleansing, monitoring, transformation, and delivery
of data, which transforms mere numbers, figures, or
character sets into reliable and consistent information.
This information can be managed in real-time, allowing
for more effective decision-making. By utilizing these
processes, fields such as academic analytics, which rely
heavily on large volumes of information, can optimize
their use of powerful data analysis tools. This allows
universities to unlock their research potential, align their
strategies with their institutional mission, and take timely
actions in pursuit of higher-quality HE.
The results of this research show that student dropout is a
multidimensional phenomenon where not only economic
factors such as social stratum, gender, type of high
school, and academic performance play an important
role but also motivation, curricular characteristics, and
sociodemographic conditions. All these factors interact
73
adoption within academic programs and increase their
effectiveness. The ADHE could update student information
on a semesterly basis to offer insights into student
attrition analysis, as certain factors may fluctuate over
time. Additionally, the ADHE could possess an in-house
data preprocessing tool to evaluate and enhance the
data quality for dependable analysis. Moreover, data
preprocessing is critical since it enables the extraction of
high-quality output information from large educational
data sources, such as the present case [2, 14].
5.2 Integration of information in HE from
different sources
Incorporating data from various sources should not
be treated as a static process, as data continues to be
generated at an increasingly rapid pace. Information
repositories that simply capture data from a single
moment in time are not effective, and integration must
be ongoing and adaptable to address challenges such
as those presented in [4], which include data velocity
and variety. As a result, it is essential that integration
processes are fast and efficient, and that the dashboards
utilized for presenting information are dynamic over time.
On the other hand, it is crucial for dashboards, as a
tool for descriptive analytics, to serve as a means of
integration for various forms of analytics. Each dashboard
undertakes an internal data processing step to present the
information in an organized, lucid, and practical manner.
However, for a dashboard to be an effective integration
tool, it should facilitate the generation of integrated
data following the internal processing step. Exporting
preprocessed raw data that can serve as input to other
analytics such as predictive and make many strategic
decisions. Unfortunately, ADHE does not include dynamic
integration or data export functions, but it is clear the
importance of them as an opportunity for future work and
research.
5.3 Characterization of student dropout in
an HEI
Dashboards and reports play a fundamental role as tools
for visualization, tracking, and information consolidation.
In this case, the ADHE allows for finding relationships
between dropouts and variables related to demographic,
social, and academic performance. From a marginal point
of view, ADHE showed there is a difference between the
dropout proportion according to social stratum, gender,
and type of high school. Also, the main dropout proportion
occurred before reaching the middle of the program.
This finding allows designing policies and projects to
analyze the phenomenon in the first semesters. Figure
6 indicates that student dropout is not solely caused by
academic factors. The figure displays that a percentage
of students who left the faculty had met the academic
requirements, having attained a cumulative GPA above
the passing grade of 3.0. Consequently, we propose that
HEI could incorporate additional factors such as stress
levels, student poverty, lack of parental involvement, and
lack of self-motivation, as these factors could heighten
the likelihood of student attrition in HE. Moreover, future
research could focus on developing predictive models for
forecasting early student dropout as well as predicting
student attrition on a semester-by-semester basis.
6. Conclusions
In this paper, we propose an academic analytic dashboard
named ADHE, which is a novel tool in academic analytics
because it allows us to integrate data from different
academic sources and depict the relations between
multiple variables interactively. ADHE can be used for
decision-makers to evaluate the relationship between
some variables and student dropout in HEI, and this new
knowledge can be used to set future policies to minimize
student dropout proportion. This visualization tool, the
result of using descriptive analytics, can also serve as a
basis for further studies on predictive and prescriptive
academic analytics. This study contributes by encouraging
HEI to build their dashboards to share useful information
with students, teachers, and administrators.
The integration of information from different sources
requires developing a series of technical and business
intelligence processes in a structured manner, given
that only through an established methodology is it
possible to convert large volumes of data - which are also
variable, heterogeneous, and generated at high velocity
- into reliable, and valuable information. An accurate
integration process facilitates the comprehension,
cleansing, monitoring, transformation, and delivery
of data, which transforms mere numbers, figures, or
character sets into reliable and consistent information.
This information can be managed in real-time, allowing
for more effective decision-making. By utilizing these
processes, fields such as academic analytics, which rely
heavily on large volumes of information, can optimize
their use of powerful data analysis tools. This allows
universities to unlock their research potential, align their
strategies with their institutional mission, and take timely
actions in pursuit of higher-quality HE.
The results of this research show that student dropout is a
multidimensional phenomenon where not only economic
factors such as social stratum, gender, type of high
school, and academic performance play an important
role but also motivation, curricular characteristics, and
sociodemographic conditions. All these factors interact
73
O. D Rivera-Baena et al., Revista Facultad de Ingeniería, Universidad de Antioquia, No. 111, pp. 64-75, 2024
in different ways and at different levels, making the
student dropout problem extremely complex. As a result,
the development of student dropout analysis requires
the involvement of various disciplines and perspectives,
including emotional, psychological, and social dimensions,
as well as factors related to academic performance, which
can be objectively measured using indicators that capture
student characteristics. Thus, it is in this context that
descriptive analytics maximizes its potential since it
allows for identifying relationships between variables,
obtaining conclusions, and improving decision-making
from objective behaviors and logical structures.
Future work may be focused on keeping the dashboard
updated, integrating the dashboard with other HEI
information sources, and including some manner of user
manipulation/selection of variables and statistical tests.
In addition, we will work on the types of intervention
according to the results obtained.
7. Declaration of competing interest
We declare that we have no significant competing interests,
including financial or non-financial, professional, or
personal interests interfering with the full and objective
presentation of the work described in this manuscript.
8. Funding
This work was supported by CODI.
9. Author contributions
Daniel Rivera Baena: Draft manuscript preparation.
Analysis and interpretation of results Carmen
Patino-Rodríguez: Data collection. Analysis and
interpretation of results. Olga Usuga-Manco: Analysis and
interpretation of results. Draft manuscript preparation
Freddy Hernández-Barajas: Study conception and design.
Analysis and interpretation of results
10. Data availability statement
The authors confirm that the data supporting
the findings of this study are available at
https://fhernanb.shinyapps.io/AppPermanencia/. The
data analyzed in this paper were taken from the data
repository of the Colombian Institute for the Evaluation of
Education, DataIcfes [33], and the institutional information
system of Universidad de Antioquia. The data analyzed
comes from three databases.
References
[1] U. Sivarajah, M. M. Kamal, Z. Irani, and V. Weerakkody, “Critical
analysis of big data challenges and analytical methods,” Journal
of Business Research, no. 70, Aug. 10, 2016. [Online]. Available:
https://doi.org/10.1016/j.jbusres.2016.08.001
[2] T. Catarci, M. Scannapieco, and M. C. and; C. Demetrescu, “My (fair)
big data,” in 2017 IEEE International Conference on Big Data (Big Data),
Boston, MA, USA, 2018.
[3] B. K. Daniel and R. Butson, “Technology enhanced analytics (tea)
in higher education,” International Association for Development
of the Information Society, Paper presented at the International
Conference on Educational Technologies, Kuala Lumpur, Malaysia,
2013.
[4] B. Daniel, “Big data and analytics in higher education: Opportunities
and challenges,” Special Issue: Open Data in Learning Technology,
vol. 46, no. 5, Dec. 22, 2014. [Online]. Available: https://doi.org/10.
1111/bjet.12230
[5] S. A.Ferreira and A. Andrade, “Academic analytics: Anatomy of an
exploratory essay,” Education and Information Technologies volume,
vol. 21, Mar. 20, 2014. [Online]. Available: https://doi.org/10.1007/
s10639-014-9317-9
[6] R. Chen, “Institutional characteristics and college student dropout
risks: A multilevel event history analysis,” Research in Higher
Education volume, vol. 53, Mar. 01, 2011. [Online]. Available:
https://doi.org/10.1007/s11162-011-9241-4
[7] M. V. López-Pérez, M. C. Pérez-López, and L. Rodríguez-Ariza,
“Blended learning in higher education: Students’ perceptions and
their relation to outcomes,” Computers & Education, vol. 56, no. 3,
Apr. 2011. [Online]. Available: https://doi.org/10.1016/j.compedu.
2010.10.023
[8] A. P. Rovai, “Sense of community, perceived cognitive learning,
and persistence in asynchronous learning networks,” The Internet
and Higher Education, vol. 5, no. 4, 2002. [Online]. Available:
https://doi.org/10.1016/S1096-7516(02)00130-6
[9] S. C. Guerrero, “Characterization of dropout in the pedagogical and
technological university of colombia during the period 2008-2015,”
Revista Lasallista de Investigación, Jun. 2018. [Online]. Available:
https://doi.org/10.22507/rli.v15n1a2
[10] Y. A. Quintero-Tangarife, “Diseño de un modelo predictivo para
generar alertas tempranas de deserción universitaria en los
programas de pregrado presenciales de la facultad de ingeniería de
la universidad de antioquia.” M.S. thesis, Fac. de Ingeniería, Univ. de
Antioquia., Medellín, Colombia, 2022.
[11] C. Parra, E. Castañeda, G. Restrepo, O. Usuga, P. Duque, and
R. Mendoza, “¿la deserción y la graduación no diferencian a los
programas de pregrado de la facultad de ingeniería de la universidad
de antioquía?” in 2014: Congreso CLABES IV., Medellin, Colombia,
2016, pp. 1–7.
[12] L. P. Navas, F. Montes, S. Abolghasem, R. J. Salas, M. Toloo, and
R. Zarama, “Colombian higher education institutions evaluation,”
Socio-Economic Planning Sciences, vol. 71, Sep. 2020. [Online].
Available: https://doi.org/10.1016/j.seps.2020.100801
[13] U. Mat, N. Buniyamin, P. Mohd-Arsad, and R. Kassim, “An overview
of using academic analytics to predict and improve students’
achievement: A proposed proactive intelligent intervention,” 2013
IEEE 5th Conference on Engineering Education (ICEED), Kuala
Lumpur, Malaysia, 2014.
[14] W. Terraza-Beleño, “Estrategias de retención estudiantil en
educación superior y su relación con la deserción revista
electrónica en educación y pedagogía,” Revista Electrónica en
Educación y Pedagogía, vol. 3, no. 4, Jun. 2019. [Online]. Available:
https://doi.org/10.15658/rev.electron.educ.pedagog19.03030403
[15] J. Géryk and L. Popelínský, “Visual analytics for increasing
efficiency of higher education institutions,” Business Information
Systems Workshops, vol. 183, Oct. 01, 2014. [Online]. Available:
https://doi.org/10.1007/978-3-319-11460-6_11
[16] P. A. Chinome-Becerra, C. Ruiz-Cárdenas, and
L. Fernández-Samacá, “Priorizacion de variables en el diseno
74
in different ways and at different levels, making the
student dropout problem extremely complex. As a result,
the development of student dropout analysis requires
the involvement of various disciplines and perspectives,
including emotional, psychological, and social dimensions,
as well as factors related to academic performance, which
can be objectively measured using indicators that capture
student characteristics. Thus, it is in this context that
descriptive analytics maximizes its potential since it
allows for identifying relationships between variables,
obtaining conclusions, and improving decision-making
from objective behaviors and logical structures.
Future work may be focused on keeping the dashboard
updated, integrating the dashboard with other HEI
information sources, and including some manner of user
manipulation/selection of variables and statistical tests.
In addition, we will work on the types of intervention
according to the results obtained.
7. Declaration of competing interest
We declare that we have no significant competing interests,
including financial or non-financial, professional, or
personal interests interfering with the full and objective
presentation of the work described in this manuscript.
8. Funding
This work was supported by CODI.
9. Author contributions
Daniel Rivera Baena: Draft manuscript preparation.
Analysis and interpretation of results Carmen
Patino-Rodríguez: Data collection. Analysis and
interpretation of results. Olga Usuga-Manco: Analysis and
interpretation of results. Draft manuscript preparation
Freddy Hernández-Barajas: Study conception and design.
Analysis and interpretation of results
10. Data availability statement
The authors confirm that the data supporting
the findings of this study are available at
https://fhernanb.shinyapps.io/AppPermanencia/. The
data analyzed in this paper were taken from the data
repository of the Colombian Institute for the Evaluation of
Education, DataIcfes [33], and the institutional information
system of Universidad de Antioquia. The data analyzed
comes from three databases.
References
[1] U. Sivarajah, M. M. Kamal, Z. Irani, and V. Weerakkody, “Critical
analysis of big data challenges and analytical methods,” Journal
of Business Research, no. 70, Aug. 10, 2016. [Online]. Available:
https://doi.org/10.1016/j.jbusres.2016.08.001
[2] T. Catarci, M. Scannapieco, and M. C. and; C. Demetrescu, “My (fair)
big data,” in 2017 IEEE International Conference on Big Data (Big Data),
Boston, MA, USA, 2018.
[3] B. K. Daniel and R. Butson, “Technology enhanced analytics (tea)
in higher education,” International Association for Development
of the Information Society, Paper presented at the International
Conference on Educational Technologies, Kuala Lumpur, Malaysia,
2013.
[4] B. Daniel, “Big data and analytics in higher education: Opportunities
and challenges,” Special Issue: Open Data in Learning Technology,
vol. 46, no. 5, Dec. 22, 2014. [Online]. Available: https://doi.org/10.
1111/bjet.12230
[5] S. A.Ferreira and A. Andrade, “Academic analytics: Anatomy of an
exploratory essay,” Education and Information Technologies volume,
vol. 21, Mar. 20, 2014. [Online]. Available: https://doi.org/10.1007/
s10639-014-9317-9
[6] R. Chen, “Institutional characteristics and college student dropout
risks: A multilevel event history analysis,” Research in Higher
Education volume, vol. 53, Mar. 01, 2011. [Online]. Available:
https://doi.org/10.1007/s11162-011-9241-4
[7] M. V. López-Pérez, M. C. Pérez-López, and L. Rodríguez-Ariza,
“Blended learning in higher education: Students’ perceptions and
their relation to outcomes,” Computers & Education, vol. 56, no. 3,
Apr. 2011. [Online]. Available: https://doi.org/10.1016/j.compedu.
2010.10.023
[8] A. P. Rovai, “Sense of community, perceived cognitive learning,
and persistence in asynchronous learning networks,” The Internet
and Higher Education, vol. 5, no. 4, 2002. [Online]. Available:
https://doi.org/10.1016/S1096-7516(02)00130-6
[9] S. C. Guerrero, “Characterization of dropout in the pedagogical and
technological university of colombia during the period 2008-2015,”
Revista Lasallista de Investigación, Jun. 2018. [Online]. Available:
https://doi.org/10.22507/rli.v15n1a2
[10] Y. A. Quintero-Tangarife, “Diseño de un modelo predictivo para
generar alertas tempranas de deserción universitaria en los
programas de pregrado presenciales de la facultad de ingeniería de
la universidad de antioquia.” M.S. thesis, Fac. de Ingeniería, Univ. de
Antioquia., Medellín, Colombia, 2022.
[11] C. Parra, E. Castañeda, G. Restrepo, O. Usuga, P. Duque, and
R. Mendoza, “¿la deserción y la graduación no diferencian a los
programas de pregrado de la facultad de ingeniería de la universidad
de antioquía?” in 2014: Congreso CLABES IV., Medellin, Colombia,
2016, pp. 1–7.
[12] L. P. Navas, F. Montes, S. Abolghasem, R. J. Salas, M. Toloo, and
R. Zarama, “Colombian higher education institutions evaluation,”
Socio-Economic Planning Sciences, vol. 71, Sep. 2020. [Online].
Available: https://doi.org/10.1016/j.seps.2020.100801
[13] U. Mat, N. Buniyamin, P. Mohd-Arsad, and R. Kassim, “An overview
of using academic analytics to predict and improve students’
achievement: A proposed proactive intelligent intervention,” 2013
IEEE 5th Conference on Engineering Education (ICEED), Kuala
Lumpur, Malaysia, 2014.
[14] W. Terraza-Beleño, “Estrategias de retención estudiantil en
educación superior y su relación con la deserción revista
electrónica en educación y pedagogía,” Revista Electrónica en
Educación y Pedagogía, vol. 3, no. 4, Jun. 2019. [Online]. Available:
https://doi.org/10.15658/rev.electron.educ.pedagog19.03030403
[15] J. Géryk and L. Popelínský, “Visual analytics for increasing
efficiency of higher education institutions,” Business Information
Systems Workshops, vol. 183, Oct. 01, 2014. [Online]. Available:
https://doi.org/10.1007/978-3-319-11460-6_11
[16] P. A. Chinome-Becerra, C. Ruiz-Cárdenas, and
L. Fernández-Samacá, “Priorizacion de variables en el diseno
74
O. D Rivera-Baena et al., Revista Facultad de Ingeniería, Universidad de Antioquia, No. 111, pp. 64-75, 2024
de un sistema de gestion integral de la desercion estudiantil,”
Revista Educacion en Ingenieria, vol. 11, no. 22, Jul-Dec. 2016.
[Online]. Available: https://tinyurl.com/5n85zt9p
[17] Y. Y. Wong, “Academic analytics: a meta-analysis of its applications
in higher education,” International Journal of Services and Standards,
vol. 11, no. 2, Jul. 26, 2016. [Online]. Available: hhttps://doi.org/10.
1504/IJSS.2016.077957
[18] L. G. Turizo-Martínez, K. García-Mendoza, S. Soto-Cantero,
Z. Fragozo-Torres, and T. J. Crissien-Borrero, “Estudio sobre
la deserción y la no graduación en la corporación universidad de
la costa, colombia,” Revista Unimar, vol. 37, no. 2, 2019. [Online].
Available: https://tinyurl.com/5328fupp
[19] E. Castañeda-Gómez, “Rendimiento académico de los estudiantes
en el primer semestre: Facultad de ingeniería cohortes 2016-1 y
2015-1,” Ingeniería y Sociedad, vol. 11, no. 1, Jan. 24, 2017. [Online].
Available: https://tinyurl.com/4r4xtc6y
[20] J. G. Villegas, C. Castañeda, and E. Castañeda-Gómez, “Planeación
y medición del desempeño en educación superior: tres casos de
aplicación de investigación de operaciones,” Ingeniería y Sociedad,
vol. 100, no. 3, Jul-Sep. 2021. [Online]. Available: https://doi.org/10.
17533/udea.redin.20210526
[21] P. Murnion and M. Helfer, “Academic analytics in quality assurance
using organisational analytical capabilities,” Uk academy for
information systems conference proceedings 2013, 2013.
[22] M. Komenda, M. Víta, C. Vaitsis, D. Schwarz, A. Pokorná, and et
al., “Curriculum mapping with academic analytics in medical and
healthcare educationr,” PLoS ONE, vol. 10, no. 12, Dec. 01, 2015.
[Online]. Available: https://doi.org/10.1371/journal.pone.0143748
[23] M. Sharkey, “Academic analytics landscape at the university of
phoenix,” in LAK ’11: Proceedings of the 1st International Conference
on Learning Analytics and Knowledge, 2011, pp. 122–126.
[24] E. J. M. Lauría, J. D. Baron, M. Devireddy, V. Sundararaju, and S. M.
Jayaprakash, “Mining academic data to improve college student
retention: an open source perspective,” in LLAK ’12: Proceedings of
the 2nd International Conference on Learning Analytics and Knowledge,
2012, pp. 139–142.
[25] E. J. M. Lauría, E. W. Moody, S. M. Jayaprakash, N. Jonnalagadda,
and J. D. Baron, “Open academic analytics initiative: initial
research findings,” in LAK ’13: Proceedings of the Third International
Conference on Learning Analytics and Knowledge, 2013, pp. 150–154.
[26] A. M. D. Rocchis, A. Michalenko, L. E. Boucheron, and S. J. Stochaj,
“Extending academic analytics to engineering education,” in 2018
IEEE Frontiers in Education Conference (FIE), San Jose, USA, 2019.
[27] L. C. Hafer, N. M. Gibson, and R. Tsemunhu, “An examination of
student retention at a 2-year college through structural equation
modeling,” Journal of College Student Retention: Research, Theory
& Practice, vol. 22, no. 4, May. 17, 2018. [Online]. Available:
https://doi.org/10.1177/1521025118770813
[28] A. M. Rodriguez, “Academic analytics: aplicando técnicas de
business intelligence sobre datos de performance académica en
enseñanza superior,” Interfaces Científicas - Exatas e Tecnológicas,
vol. 1, no. 2, May. 28, 2015. [Online]. Available: https://doi.org/10.
17564/2359-4942.2015v1n2p35-46
[29] C. Shi, S. Fu, Q. Chen, and H. Qu, “Vismooc: Visualizing video
clickstream data from massive open online courses,” 2015 IEEE
Pacific Visualization Symposium (PacificVis), Hangzhou, 2015.
[30] H. He, O. Zheng, and B. Dong, “Vusphere: Visual analysis of video
utilization in online distance education,” 2018 IEEE Conference on
Visual Analytics Science and Technology (VAST), Berlin, Germany,
2018.
[31] Y. Chen, Q. Chen, M. Zhao, S. Boyer, K. Veeramachaneni, and
H. Qu, “Dropoutseer: Visualizing learning patterns in massive open
online courses for dropout reasoning and prediction,” 2016 IEEE
Conference on Visual Analytics Science and Technology (VAST),
Baltimore, USA, 2017.
[32] M. McNaughton, L. Rao, and G. Mansingh, “An agile approach for
academic analytics: a case study,” Journal of Enterprise Information
Management, vol. 30, no. 5, Sep. 11, 2017. [Online]. Available:
http://www.techweb.com/se/index.html
[33] G. de Colombia. Icfes. Accessed 2020. [Online]. Available: https:
//www.icfes.gov.co/acceso-a-bases-de-datos-y-diccionarios
[34] J. A. Y. X. W. Chang, J. Cheng and J. McPherson. shiny: Web
Application Framework for R. Accessed 2020. [Online]. Available:
https://shiny.posit.co/r/reference/shiny/1.4.0/shiny-package.html
[35] W. Chang and B. Borges-Ribeiro. shinydashboard: Create
Dashboards with Shiny. Accessed 2019. [Online]. Available: https:
//cran.r-project.org/web/packages/shinydashboard/index.html
[36] V. Perrier, F. Meyer, and D. Granjon. shiny Widgets: Custom
Inputs Widgets for Shiny. Accessed 2020. [Online]. Available:
https://dreamrs.github.io/shinyWidgets/
[37] H. Wickham, ggplot2: Elegant Graphics for Data Analysis. New York:
Springer-Verlag, 2016.
[38] C. Sievert. (2020) Interactive web-based data visualization with
r, plotly, and shiny. Chapman and Hall/CRC. [Online]. Available:
https://tinyurl.com/56mve93p
[39] H. Wickham, R. François, L. Henry, and K. Müller. (2019, Feb.)
A grammar of data manipulation. [Online]. Available: http://dplyr.
tidyverse.org,https://github.com/tidyverse/dplyr
[40] J. C. Y. Xie and X. Tan. DT: A Wrapper of the JavaScript
Library ’DataTables’. Accessed 2020. [Online]. Available: https:
//cran.r-project.org/web/packages/DT/index
[41] S. Garnier. viridis: Default Color Maps from ’matplotlib. Accessed
2019. [Online]. Available: https://cran.r-project.org/web/packages/
viridis/index.html
75
de un sistema de gestion integral de la desercion estudiantil,”
Revista Educacion en Ingenieria, vol. 11, no. 22, Jul-Dec. 2016.
[Online]. Available: https://tinyurl.com/5n85zt9p
[17] Y. Y. Wong, “Academic analytics: a meta-analysis of its applications
in higher education,” International Journal of Services and Standards,
vol. 11, no. 2, Jul. 26, 2016. [Online]. Available: hhttps://doi.org/10.
1504/IJSS.2016.077957
[18] L. G. Turizo-Martínez, K. García-Mendoza, S. Soto-Cantero,
Z. Fragozo-Torres, and T. J. Crissien-Borrero, “Estudio sobre
la deserción y la no graduación en la corporación universidad de
la costa, colombia,” Revista Unimar, vol. 37, no. 2, 2019. [Online].
Available: https://tinyurl.com/5328fupp
[19] E. Castañeda-Gómez, “Rendimiento académico de los estudiantes
en el primer semestre: Facultad de ingeniería cohortes 2016-1 y
2015-1,” Ingeniería y Sociedad, vol. 11, no. 1, Jan. 24, 2017. [Online].
Available: https://tinyurl.com/4r4xtc6y
[20] J. G. Villegas, C. Castañeda, and E. Castañeda-Gómez, “Planeación
y medición del desempeño en educación superior: tres casos de
aplicación de investigación de operaciones,” Ingeniería y Sociedad,
vol. 100, no. 3, Jul-Sep. 2021. [Online]. Available: https://doi.org/10.
17533/udea.redin.20210526
[21] P. Murnion and M. Helfer, “Academic analytics in quality assurance
using organisational analytical capabilities,” Uk academy for
information systems conference proceedings 2013, 2013.
[22] M. Komenda, M. Víta, C. Vaitsis, D. Schwarz, A. Pokorná, and et
al., “Curriculum mapping with academic analytics in medical and
healthcare educationr,” PLoS ONE, vol. 10, no. 12, Dec. 01, 2015.
[Online]. Available: https://doi.org/10.1371/journal.pone.0143748
[23] M. Sharkey, “Academic analytics landscape at the university of
phoenix,” in LAK ’11: Proceedings of the 1st International Conference
on Learning Analytics and Knowledge, 2011, pp. 122–126.
[24] E. J. M. Lauría, J. D. Baron, M. Devireddy, V. Sundararaju, and S. M.
Jayaprakash, “Mining academic data to improve college student
retention: an open source perspective,” in LLAK ’12: Proceedings of
the 2nd International Conference on Learning Analytics and Knowledge,
2012, pp. 139–142.
[25] E. J. M. Lauría, E. W. Moody, S. M. Jayaprakash, N. Jonnalagadda,
and J. D. Baron, “Open academic analytics initiative: initial
research findings,” in LAK ’13: Proceedings of the Third International
Conference on Learning Analytics and Knowledge, 2013, pp. 150–154.
[26] A. M. D. Rocchis, A. Michalenko, L. E. Boucheron, and S. J. Stochaj,
“Extending academic analytics to engineering education,” in 2018
IEEE Frontiers in Education Conference (FIE), San Jose, USA, 2019.
[27] L. C. Hafer, N. M. Gibson, and R. Tsemunhu, “An examination of
student retention at a 2-year college through structural equation
modeling,” Journal of College Student Retention: Research, Theory
& Practice, vol. 22, no. 4, May. 17, 2018. [Online]. Available:
https://doi.org/10.1177/1521025118770813
[28] A. M. Rodriguez, “Academic analytics: aplicando técnicas de
business intelligence sobre datos de performance académica en
enseñanza superior,” Interfaces Científicas - Exatas e Tecnológicas,
vol. 1, no. 2, May. 28, 2015. [Online]. Available: https://doi.org/10.
17564/2359-4942.2015v1n2p35-46
[29] C. Shi, S. Fu, Q. Chen, and H. Qu, “Vismooc: Visualizing video
clickstream data from massive open online courses,” 2015 IEEE
Pacific Visualization Symposium (PacificVis), Hangzhou, 2015.
[30] H. He, O. Zheng, and B. Dong, “Vusphere: Visual analysis of video
utilization in online distance education,” 2018 IEEE Conference on
Visual Analytics Science and Technology (VAST), Berlin, Germany,
2018.
[31] Y. Chen, Q. Chen, M. Zhao, S. Boyer, K. Veeramachaneni, and
H. Qu, “Dropoutseer: Visualizing learning patterns in massive open
online courses for dropout reasoning and prediction,” 2016 IEEE
Conference on Visual Analytics Science and Technology (VAST),
Baltimore, USA, 2017.
[32] M. McNaughton, L. Rao, and G. Mansingh, “An agile approach for
academic analytics: a case study,” Journal of Enterprise Information
Management, vol. 30, no. 5, Sep. 11, 2017. [Online]. Available:
http://www.techweb.com/se/index.html
[33] G. de Colombia. Icfes. Accessed 2020. [Online]. Available: https:
//www.icfes.gov.co/acceso-a-bases-de-datos-y-diccionarios
[34] J. A. Y. X. W. Chang, J. Cheng and J. McPherson. shiny: Web
Application Framework for R. Accessed 2020. [Online]. Available:
https://shiny.posit.co/r/reference/shiny/1.4.0/shiny-package.html
[35] W. Chang and B. Borges-Ribeiro. shinydashboard: Create
Dashboards with Shiny. Accessed 2019. [Online]. Available: https:
//cran.r-project.org/web/packages/shinydashboard/index.html
[36] V. Perrier, F. Meyer, and D. Granjon. shiny Widgets: Custom
Inputs Widgets for Shiny. Accessed 2020. [Online]. Available:
https://dreamrs.github.io/shinyWidgets/
[37] H. Wickham, ggplot2: Elegant Graphics for Data Analysis. New York:
Springer-Verlag, 2016.
[38] C. Sievert. (2020) Interactive web-based data visualization with
r, plotly, and shiny. Chapman and Hall/CRC. [Online]. Available:
https://tinyurl.com/56mve93p
[39] H. Wickham, R. François, L. Henry, and K. Müller. (2019, Feb.)
A grammar of data manipulation. [Online]. Available: http://dplyr.
tidyverse.org,https://github.com/tidyverse/dplyr
[40] J. C. Y. Xie and X. Tan. DT: A Wrapper of the JavaScript
Library ’DataTables’. Accessed 2020. [Online]. Available: https:
//cran.r-project.org/web/packages/DT/index
[41] S. Garnier. viridis: Default Color Maps from ’matplotlib. Accessed
2019. [Online]. Available: https://cran.r-project.org/web/packages/
viridis/index.html
75