DOI: https://doi.org/
1
0.56712/latam.v7i4.6409
Predicting the use of learning objects: supporting high school
education
Predicción del uso de objetos de aprendizaje: apoyo a la educación
secundaria
Adriana Pérez López
adriana.pl@teziutlan.tecnm.mx
https://orcid.org/0000-0003-3712-400X
Instituto Tecnológico Superior de Teziutlán /TecNM
Teziutlán – México
Raúl Mora Reyes
raul.mr@teziutlan.tecnm.mx
https://orcid.org/0000-0002-1428-8901
Instituto Tecnológico Superior de Teziutlán/TecNM
Teziutlán – México
Naty Rodríguez Ventura
naty.rv@teziutlan.tecnm.mx
https://orcid.org/0009-0007-5976-4418
Instituto Tecnológico Superior de Teziutlán/TecNM
Teziutlán – México
Sara Aros Alberto
M23TE0050@teziutlan.tecnm.mx
https://orcid.org/0009-0004-4116-499X
Instituto Tecnológico Superior de Teziutlán/TecNM
Teziutlán – México
Artículo recibido: 03 de abril 2026. Aceptado para publicación: 03 de septiembre de 2026.
Conflictos de Interés: Ninguno que declarar.
Abstract
Education today faces major challenges compared to the situation before 2020, due to changes in the
ways of teaching and learning caused by the difficulties of attending classes during the COVID
-
19
pandemic.
Among
the
main
challenges
are
the
lack
of
didact
ic
resources,
limited
technological
infrastructure, and the geographical location of many schools, particularly those in rural areas. This
research analyzes the effectiveness of a predictive model based on the Random Forest algorithm to
suggest personalize
d learning objects for high school students, through the analysis of academic and
behavioral
variables
such
as
frequency
and
time
of
use
of
digital
resources,
motivation,
and
the
perceived impact on comprehension. The findings confirm that the predictive m
ethodology not only
supports educational personalization but also generated recommendations tailored to each student’s
profile,
optimizing
the
use
of
digital
resources
and
significantly
improving
comprehension
and
academic
performance.
Validation,
carried
out
through
a
diagnostic
pre
-
test
and
a
final
post
-
test,
demonstrated
a
significant
average
improvement
of
45.24%
in
the
academic
performance
of
the
experimental group that utilized the personalized learning objects suggested by the model. This study
propo
ses
the
integration
of
machine
learning
models
as
an
adaptive
strategy
in
educational
environments, highlighting the capacity of data analytics to promote equity, reduce learning disparities,
and strengthen student
-
centered pedagogy.
Keywords:
learning objects, predictive model, academic performance, student motivation
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2026, Volumen VII, Número 4 p 2433.
Resumen
La educación se enfrenta hoy en día a importantes retos en comparación con la situación anterior a
2020, debido a los cambios en las formas de enseñar y aprender provocados por las dificultades para
asistir a clase durante la pandemia de COVID-19. Entre los principales retos se encuentran la falta de
recursos
didácticos,
la
infraestructura
tecnológica
limitada
y
la
ubicación
geográfica
de
muchos
centros educativos, especialmente los situados en zonas rurales. Esta investigación analiza la eficacia
de un modelo predictivo basado en el algoritmo Random Forest para sugerir objetos de aprendizaje
personalizados
a
alumnos
de
secundaria,
mediante
el
análisis
de
variables
académicas
y
de
comportamiento, como la frecuencia y el tiempo de uso de los recursos digitales, la motivación y el
impacto percibido en la comprensión. Los resultados confirman que la metodología predictiva no solo
favorece la personalización educativa, sino que también generó recomendaciones adaptadas al perfil
de
cada
alumno,
optimizando
el
uso
de
los
recursos
digitales
y
mejorando
significativamente
la
comprensión
y
el
rendimiento
académico.
La
validación,
llevada
a
cabo
mediante
una
prueba
de
diagnóstico y un examen final, mostró una mejora media del 45,24 % en el rendimiento académico del
grupo
experimental
que
utilizó
los
objetos
de
aprendizaje
sugeridos
por
el
modelo.
Este
estudio
propone la integración de modelos de aprendizaje automático como estrategia adaptativa en entornos
educativos,
destacando
la
capacidad
del
análisis
de
datos
para
promover
la
equidad,
reducir
las
disparidades en el aprendizaje y reforzar la pedagogía centrada en el alumno.
Palabras clave:
objetos de aprendizaje, modelo predictivo, rendimiento académico, motivación
de los alumnos
Todo el contenido de LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades,
publicado en este sitio está disponibles bajo Licencia
Creative Commons
.
Cómo citar: Pérez López, A., Mora Reyes, R., Rodríguez Ventura, N., & Aros Alberto, S. (2026).
Predicting the use of learning objects: supporting high school education
.
LATAM Revista
Latinoamericana de Ciencias Sociales y Humanidades 7 (4), 2433 – 2452.
https://doi.org/10.56712/latam.v7i4.6409
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2026, Volumen VII, Número 4 p 2434.
INTRODUCTION
In 2015, the United Nations General Assembly approved the 2030 Agenda, which is composed of 17
Sustainable
Development
Goals.
In
Mexico,
the
three
goals
with
the
least
progress
are:
Goal
9
–
Industry, Innovation, and Infrastructure; Goal 15
–
Life on Land;
and Goal 4
–
Quality Education (Global
Compact, 2025). The last one is where the use of Learning Objects (LO) is being promoted. According
to data from the National Institute of Statistics and Geography (INEGI) and the Secretariat of Public
Education
(SEP
),
during
the
2022
–
2023
period,
the
graduation
efficiency
rate
in
upper
secondary
education in the states of Puebla and Veracruz was 77.2% and 75.6%, respectively (INEGI, 2025).
The predominant factor in these figures is school dropout, which in both states during the same period
was 8.7%. The main reasons include: financial difficulties that force students to work, low academic
performance and failure, lack of interest in studies
, and gaps in prior educational quality (Hernández
Osorio, 2023).
Against this backdrop, and based on data from the 2023 National Survey on Availability and Use of
Information Technologies in Households (National Institute of Statistics and Geography, 2023), internet
usage among youth aged 12 to 17 in rural areas of Mexi
co reached 92.4%, mostly through mobile
devices.
This
brings
additional
benefits
such
as:
easier
access
to
digital
resources,
promotion
of
knowledge
exchange,
expansion
of
educational
horizons,
and
implementation
of
innovative
pedagogical strategies. There
fore, this research focuses on upper secondary schools, including both
preparatory and high school institutions (Serna Martínez & Alvites Huamaní, 2021).
This project explores the conceptual foundations that support the use of learning objects, with a focus
on their application in rural areas of Teziutlan, a region in the state of Puebla. The study builds upon
previous research that evaluated the use of lea
rning objects in educational contexts characterized by
limited resources and poor learning conditions. These factors are considered indicators for assessing
the practical effectiveness of learning objects, their impact on academic performance, and their va
lue
as pedagogical tools for educators within the teaching and learning process. (Orellana
-
Cordero et al.,
2020) The aim of this study is to establish criteria for determining whether learning objects effectively
contribute to achieving educational objecti
ves in such environments. (Aros, 2025).
METHODOLOGY
Research approach
The
study
is
based
on
a
quantitative,
predictive,
and
cross
-
sectional
approach,
supported
by
the
analysis
of
historical
data
on
the
use
of
Learning
Objects
(LO)
among
higher
education
students
(Palmera
Quintero
et
al.,
2023).
This
methodological
design
was
chosen
for
its
ability
to
identify
behavioral patterns and generate predictive models that help optimize pedagogical strategies.
To carry out this study, the CRISP
-
DM methodology was implemented
—
commonly used in data mining
projects due to its adaptability to various educational contexts. It consists of six phases (Hotz, 2024),
described as follows:
Business Understanding:
This phase focuses on ensuring alignment between the educational objectives of the project and the
proposed data analysis, guaranteeing that decisions address real needs within the academic context.
Educational
environment
needs
were
identified,
and
key
qu
estions
were
formulated
to
guide
the
analysis.
The
main
goal
of
this
phase
is
to
predict
which
LOs
show
the
highest
levels
of
use
and
acceptance among high school students in the Teziutlán region, in order to personalize teaching and
improve academic perfo
rmance.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2026, Volumen VII, Número 4 p 2435.
To achieve this, activities such as the following: structured interviews and meetings with teachers,
coordinators,
and
principals
to
understand
their
needs
and
identify
pedagogical
and
technological
gaps. Key questions were formulated, such as: What impact
would identifying the most
-
used LOs have?
How would the results influence educational planning?, Which LO are considered most necessary to
reinforce learning?, Analytical questions were also designed, including: What usage patterns do LO
exhibit?; Are cer
tain LO more commonly used in specific subjects?, What factors influence the selection
of LO? Study metrics were defined, including: frequency of use for each LO, average interaction time
per LO, and preferences by subject, group, or academic level.
This phase also required contextual analysis, evaluating the technological infrastructure of schools and
available data sources (surveys, records, logs), while recognizing limitations such as lack of historical
data or inconsistencies. A timeline was estab
lished based on CRISP
-
DM phases, with role assignments
for data collection, model development, and validation. Necessary tools were identified: Python, MySQL,
Jupyter Notebook, Power BI.
Data Understanding
After data collection, dependent and independent variables were identified to address the research
problem.
The
dataset
included
information
such
as
age,
gender,
and
predominant
learning
styles.
Additional
questions
were
posed
to
gather
insights:
How
motiv
ated
do
you
feel
using
this
type
of
learning?, How many hours do you spend studying or reviewing school content?, How often do you
access platforms containing Learning Objects (e.g., Kahoot, Duolingo, videos)?, What is the average
time you dedicate to eac
h session?, What type of LO is most used?, Which LO do you prefer to use?,
How often do you complete the LO?, Do you consider LO to be effective?, Are LO interactive?, Do LO
help improve your academic performance?, Table 1 presents the description of the v
ariables used.
Table 1
Description of study variables
|
NOMBRE_VARIABLE
|
TIPO_DATO
|
DESCRIPTION
|
|
Edad
|
Numeric (Integer)
|
Age of students
|
|
Género
|
Nominal categorical
|
Gender of students (female, male, other)
|
|
Grado
|
Ordinal categorical
|
Grade level of students (1st, 2nd, 3rd)
|
|
acceso_internet
|
Binary categorical
|
Indicate whether the student has access to the
internet.
|
|
utiliza_OA
|
Nominal categorical
|
If you have ever used LO
|
|
frecuencia
|
Ordinal categorical
|
Frequency of LO use (sometimes, never,
frequently, etc.)
|
|
tiempo_promedio
|
Ordinal categorical
|
Average time spent by students on LO
|
|
OA_frecuencia
|
Ordinal categorical
|
Frequency with which students use LO
|
|
OA_mayorfrecuencia
|
Nominal categorical
|
Type of LO used by the student (videos, crossword
puzzles, etc.)
|
|
OA_mejora
|
Binary categorical
|
If you have seen any improvement when using LO
|
|
OA_comprensión
|
Binary categorical
|
Whether the student understood the content of the
learning objectives.
|
|
OA_rendimiento
|
Binary categorical
|
If LO helped you improve your academic
performance
|
|
OA_barrera
|
Binary categorical
|
The bars presented by the student, in order to use
the LO
|
|
OA_motivacion
|
Binary categorical
|
If the student feels motivated when using LO
|
Source:
own elaboration.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2026, Volumen VII, Número 4 p 2436.
Data Preparation
Once the procedure was completed and based on the data provided in the questions, the analysis was
segmented as shown in Figure 1. It begins with data collection, a phase in which records related to OA
use
are
obtained
through
surveys.
Next,
the
data
is
de
scribed,
identifying
its
characteristics
and
structure, as well as the most important variables. This is followed by the data exploration phase, where
patterns, trends, and relationships between the most important variables are analyzed. Finally, data
qual
ity verification is carried out to ensure that the information used is reliable and suitable for the
predictive model, which details the data collection activities.
Data preparation process that is part of the CRISP
-
DM methodology to ensure that the information used
is reliable and suitable for the predictive model.
Figure 1
Tasks for data activities
Source:
own elaboration.
Data Collection
Data was obtained through surveys, previous studies, and academic grades by group or subject. Key
variables were identified, such as age, gender, grade level, internet access, usage frequency, motivation,
and comprehension. Null or irrelevant variables wer
e removed, and the target variable was defined: the
most frequently used type of learning object.
Data Description
Key variables necessary to build the predictive model were identified and selected, with the aim of
improving the accuracy and effectiveness of the analysis.
The variables were obtained through a survey, allowing for the collection of representative data from
the analyzed sample. Figure 2 presents the variables considered in the model, along with a detailed
description of each one. Extract from the database use
d in the predictive analysis showing the main
variables of the predictive model applied in the research, which allowed us to identify behavior patterns
and establish the relationship between OA use and academic performance.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2026, Volumen VII, Número 4 p 2437.
Figure 2
Extracted variables
Source:
own elaboration.
Data Exploration
The data was filtered to include only high school students. Columns were renamed to facilitate analysis
in
Python.
Duplicates
were
removed
and
formats
were
standardized.
Variables
such
as
age
were
transformed to numerical format and type_oa_frecuente to bi
nary format using MultiLabelBinarizer.
Similar
subjects
were
grouped,
metrics
such
as
average
usage
time
were
calculated,
and
multiple
responses were encoded in binary columns. The dataset was prepared for predictive analysis.
Table 2
Renaming of data table
|
Name
|
Renamed
|
|
timestamp
|
fecha_respuesta
|
|
Age
|
edad
|
|
Gender
|
genero
|
|
Grade level
|
grado
|
|
Do you have internet access at home?
|
acceso_internet
|
|
Have you used online learning objects?
|
uso_objetos_aprendizaje
|
|
If yes, how often do you use them?
|
frecuencia_uso
|
|
How long, on average, do you use these resources per session?
|
tiempo_promedio
|
|
What type of learning objects do you use most often?
|
tipo_oa_frecuente
|
|
Where do you access learning objects most often?
|
lugar_acceso
|
|
Do you feel that using learning objects has improved your
understanding of the topics?
|
impacto_comprension
|
|
How easy do you find it to understand the content of learning objects?
|
facilidad_comprension
|
|
How do you think these resources influence your academic
performance?
|
impacto_calificacion
|
|
What is the main barrier to using learning objects in your case?
|
barrera_uso
|
|
Do you feel more motivated to study when you use learning objects?
|
motivacion_uso
|
Source:
own elaboration.
Data quality verification
Statistical and visual techniques were applied to understand the structure of the data. Boxplots were
generated to compare age vs. objective, where it was observed whether there was any difference in the
age distribution between those who use and those who
do not use educational videos. This information
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2026, Volumen VII, Número 4 p 2438.
is shown in graphic 1. Box plot identifying that the use of educational videos is not concentrated in a
specific age range.
Graphic 1
Boxplot Age vs objective
Source:
own elaboration.
Through the creation of the heat map, graphic 2, it was possible to observe the correlation between the
predictor
variables
and
the
target
variable,
thereby
detecting
the
pattern
of
relationship
between
variables.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2026, Volumen VII, Número 4 p 2439.
Graphic 2
Predictive variables and objective
Source:
own elaboration.
Heat map of correlations, where shades closer to red show stronger positive correlations, while blues
represent negative correlations.
Motivation and impact were evaluated using a scatter plot, which allowed the relationship between both
factors to be visualized clearly and accurately, facilitating the identification of patterns, trends, and
possible correlations between the participants'
level of motivation and the impact generated in the
context of the study. These data are shown in graphic 3. The scatter plot showed that students with
higher motivation tend to indicate a positive impact on their understanding.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2026, Volumen VII, Número 4 p 2440.
Graphic 3
Motivation and Impacts of LO
Source:
own elaboration.
Lastly, within this phase, a histogram was used to analyze frequency and usage in order to identify who
does and does not use videos as a study tool. These results are shown in graphic 4. Binary distribution
of the object variable, which is related to the
use of educational videos as a learning object. It can be
seen that most students (value 1) do use educational videos, while a smaller group (value 0) say they
do not use them.
Graphic 4
Use of educational videos
Source:
own elaboration.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2026, Volumen VII, Número 4 p 2441.
D. Modeling
In this phase, predictive analysis models were developed to identify patterns in the data and anticipate
which learning objects are most used by students. The purpose is to generate useful information for
educational decision
-
making.
To
achieve
an
effective
model,
key
questions
were
raised
regarding
algorithm
selection,
variable
transformation, data division, performance evaluation, and model scalability.
Data preparation and division
Historical data collected between December 2024 and April 2025 was used; the target variable was:
most used type of learning object, supervised machine learning with multi
-
class classification was
applied, and categorical variables were transformed into nu
merical variables using one
-
hot encoding.
Missing values were imputed using the most frequent value replacement technique.
Division of the dataset
70% for model training; 10% for validation and hyperparameter tuning; and 20% for final model testing
with unseen data.
E. Evaluation
During
this
phase,
the
performance
of
the
Random
Forest
model
was
evaluated.
This
model
was
designed to recommend learning objects to high school students with low performance in mathematics
and English.
Cross
-
validation and test set tools were applied to ensure generalization of results. The metrics applied
were:
accuracy:
overall
percentage
of
correct
predictions;
precision
and
recall:
quality
of
positive
predictions; F1
-
score: balance between precision
and recall; and confusion matrix: visual identification
of hits and misses by class.
The results of this phase demonstrated the model's high performance in predicting and categorizing
student profiles; Random Forest showed resistance to overfitting and good handling of large volumes
of variables. With the application of feature importance
analysis, it was possible to identify the student
attributes most
influential
in the predictions, graphic
5 shows the
results
of two of the
algorithms
mentioned. Comparative results of three machine learning algorithms: Logistic Regression, Random
Forest,
and KNN, including evaluation metrics and confusion matrices.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2026, Volumen VII, Número 4 p 2442.
Graphic 5
Evaluation of results
Source:
own elaboration.
The model met the objectives set out in the planning phase and is therefore considered viable for
practical application, although the possibility of making iterative adjustments remains open if faults are
detected in specific subgroups (e.g., improving dat
a or adjusting hyperparameters).
F. Implementation
In
this
final
stage,
the
models
developed
are
implemented
in
a
real
environment
with
the
aim
of
translating their predictions and findings into concrete actions. The purpose is to ensure that the results
of the analysis do not remain theoretical, but are i
ntegrated into educational decision
-
making.
Among the practical applications of the model, teachers receive reports that help them identify the
most effective resources; educational platforms are updated with the most relevant learning objects;
and automated systems can monitor the use of these reso
urces in real time.
To
visualize
the
results,
an
interactive
tool
was
developed
using
Power
BI,
which
allowed
for
the
creation of graphs, maps, and dynamic tables; intuitive filtering and segmentation of data; connection
to multiple data sources (SQL, Excel, APIs); configurat
ion of automatic updates; timely detection of
trends or problems; and provision of an accessible and cost
-
effective solution,
especially useful in
educational environments.
The impact of deploying the predictive model based on historical data made it possible to identify the
most
frequently
used
learning
objects,
revealing
patterns
of
preference
among
students.
This
translates into concrete benefits for teachers, administrato
rs, and students, facilitating class planning
with more effective resources and supporting data
-
driven strategic decisions by ensuring access to
more useful and personalized materials, respectively.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2026, Volumen VII, Número 4 p 2443.
To sum up, the deployment not only displays statistics, but also transforms data into practical tools
that improve the educational experience. Interactive visualizations allow for dynamic exploration of
information,
promoting
informed
decisions
and
continu
ous
improvements
in
the
teaching
-
learning
process.
Population and sample
The study population focused on high schools in the Teziutlán region, with 1,250 students from eight
institutions, all selected through non
-
probabilistic convenience sampling, which included 22 students
from the Henry Wallon A.C. Institute, (HW) divided in
to two groups, one control and one experimental,
who were given general OAs and OAs recommended by the model, respectively.
To validate the statistical significance of the improvement observed in the experimental group (Pre
-
test
mean:
10.73,
SD:
2.86;
Post
-
test
mean:
15.27,
SD:
2.10),
a
paired
-
samples
t
-
test
was
conducted(Triola F. Mario, 2004). The results revealed a statistic
ally significant increase in academic
performance ($t(10) = 6.42, p < 0.001$), confirming that the personalized recommendations derived
from the Random Forest model generated a real pedagogical effect rather than a random variation
(Rodríguez, 2023).
DEVELOPMENT
There are cases documented in educational literature where technology has been key to significant
improvements
in
the
teaching
and
learning
process.
Such
is
the
case
of
research
conducted
in
Pakistan, which sought to address the lack of studies focused on
that country's specific context with
regard to global educational trends. While global trends are valuable, it is essential to understand how
these
innovations
and
changes
manifest
themselves
in
the
educational
environment
of
Pakistan,
specifically
in
Punj
ab.
The
main
objective
of
the
study
is
to
thoroughly
analyze
the
applicability,
difficulties, and possible modifications of global educational trends and innovations in the sociocultural
and educational environment of Pakistan. The study focuses on the imp
act of technological advances,
pedagogical
changes,
inclusive
education
practices,
and
the
globalization
of
education
to
provide
information
that
can
guide
educational
policies
and
practices
in
Pakistan.
A
quantitative
research
design was used. With a coho
rt size of 160 teachers from Punjab, simple random sampling was used
to ensure that each teacher had an equal chance of being selected. Data collection was carried out
through a self
-
developed questionnaire that was pre
-
tested to ensure clarity and reliabi
lity, using face
-
to
-
face interviews. The technologies applied include Artificial Intelligence (AI), with adaptive learning
platforms to personalize instructional content according to students' needs, and Virtual Reality (VR),
providing
immersive
learning
e
xperiences,
such
as
simulations,
that
exceed
the
capabilities
of
traditional classrooms. As a result, the vast majority of participants showed a favorable attitude toward
the integration of technology in classrooms (81% agree), the effectiveness of student
-
centered learning
approaches
(73%
agree),
and
the
relevance
of
global
educational
trends
in
the
local
context
(64%
agree). Similarly, there were divergent opinions on the challenges related to technological infrastructure
and the importance of cultural co
mpetence as a key factor in promoting inclusion. (Hadayat et al., 2024)
Another case is mentioned by (De Medio et al., 2019) in their publication on the use of MoodleRec for
course creation via Moodle platforms, the web offers a wide variety of digital resources for learning
environments, available to both teachers and student
s.
In the book Learning Objects as Support in the Teaching
-
Learning Process (Tovar et al,.2019), Chapter
6 discusses a case at the Colegio de Postgraduates. The issue identified was that despite progress in
integrating
Information
and
Communication
Technologi
es
(ICTs)
into
higher
and
postgraduate
education, there remains a need to design LOs that are efficient, reliable, adaptable, and open. The need
which is presented in the school before mentioned was to strengthen digital libraries with educational
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2026, Volumen VII, Número 4 p 2444.
resources
aligned
to
international
standards
and
solid
pedagogical
frameworks.
The
ADDIE
model
(Analysis, Design, Development, Implementation, and Evaluation) was used in the project, along with
technological tools such as PHP language, web servers, and me
tadata managers. The result was the
design and development of a specific LO on the entity
-
relationship model, aimed at students learning
database management. It was evaluated using formal instruments and a complete metadata proposal
based on IEEE
-
LOM. This
work demonstrated that the developed resources can be used freely by higher
education teachers and students under open licenses.
Similarly, (Pulido et al. 2016), in their work on LO as educational resources in algorithm instruction,
studied 47 students from Group 301 of the Bachelor's in Administrative Computer Systems (LSCA) at
the Faculty of Accounting and Administration, Universi
dad Veracruzana, Coatzacoalcos campus, during
the August 2014
–
February 2015 period. The goal was to address challenges in the LSCA program’s
Algorithmics course,
where students struggled with
algorithmic structures such as if
-
then, for, and
while. This l
ed to the design and implementation of LOs as didactic tools to improve understanding of
these complex topics. The project used a quantitative, non
-
experimental, cross
-
sectional methodology,
with
an
11
-
item
questionnaire
(2
demographic
and
9
specific
items
).
Variables
evaluated
included
content,
aesthetics,
and
functionality.
The
questionnaire
was
validated
through
a
pilot
test
(10
students) and Cronbach’s alpha for reliability (>0.7). Results of the three variables were satisfactory:
Content: 86% said the
LO information was complete; 92% found the content aligned with the curriculum;
78% said it facilitated learning. Aesthetics: 96% found the colors appropriate; 90% said the typography
was legible; 70% found the images suitable, though 23% noted some were t
oo large, identifying figure
size as an area for improvement. Functionality: 76% found the LO interaction clear, but 82% perceived
slow response speed, indicating a need for improvement.
In conclusion, the applied LOs proved effective in terms of content and learning facilitation. However,
aspects
such
as
performance
speed
and
visual
presentation
need
enhancement.
Overall
student
perception was positive, suggesting high potential for this
type of resource as a didactic tool in higher
education.
This study focuses on the use of data science techniques to improve high school Teziutlán, Puebla City.
Specifically, it proposes
a predictive model based
on the Random Forest algorithm to recommend
personalized
LOs
to
high
school
students,
aiming
to
impro
ve
academic
performance,
especially
in
critical subjects such as mathematics and English. This approach is grounded in the Cross Industry
Standard
Process
for
Data
Mining
(CRISP
-
DM),
ensuring
a
systematic
process
for
data
collection,
analysis, and modeling
. The study emerges in response to the educational
lag exacerbated by the
COVID
-
19 pandemic, which highlighted the need for effective and adaptive digital resources in rural
contexts.
RESULTS
The analysis was carried out by carefully cleaning and transforming the data collected in a survey
administered
to
high
school
students
at
HW
in
Teziutlán,
Puebla.
Categorical
variables
were
transformed into numerical variables using one
-
hot encoding (pd.g
et_dummies()), which allowed them
to be used in machine learning models. In addition, missing values were completed using the mode
strategy with SimpleImputer, ensuring the consistency of the dataset. This is shown in Figure 8 in the
code written in Python
; code snippet written in Python that is used to prepare data in the predictive
model. This process ensures that the data is ready to train the algorithms.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2026, Volumen VII, Número 4 p 2445.
Figure 3
Encoding techniques
Source:
own elaboration.
The target variable of the predictive classification problem was operationalized as the preference and
frequent use of specific Learning Objects (specifically Educational Videos, binarized as 1 = Yes, 0 = No
via MultiLabelBinarizer encoding), based on beha
vioral and motivational predictors.
Table 3.
Predictor variables
When training and comparing the three supervised classification algorithms, it was observed that the
results were very similar in performance, which shows the consistency of the patterns found in the data
and the robustness of the target variable.
Details a comparison between each of the models, which provides a broader overview for making the
best decision about which algorithm to train
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2026, Volumen VII, Número 4 p 2446.
Graphic 6
Comparison of predictive models by metric
Source:
own elaboration.
Although
Logistic
Regression
achieved
slightly
higher
global
metrics
(Accuracy:
0.729,
F1:
0.837),
Random Forest was selected as the core algorithm for the recommendation engine (Accuracy: 0.675,
Recall: 0.906, F1: 0.798). This choice is justified by Rando
m Forest's superior ability to model non
-
linear
interactions among heterogeneous educational variables, its intrinsic resistance to overfitting, and its
feature importance capability, which provides actionable pedagogical insights to teachers regarding
whi
ch
factors
(e.g.,
student
motivation,
session
duration)
most
strongly
dictate
resource
adoption;
Table 3.
Table 3
Results of Predictive Models
|
Algorithm
|
Precision
|
Recall
|
F1-Score
|
Accuracy
|
|
Logistic Regression
|
0.729
|
0.982
|
0.837
|
0.729
|
|
Random Forest (Selected)
|
0.713
|
0.906
|
0.798
|
0.675
|
|
K-Nearest Neighbors (KNN)
|
0.709
|
0.847
|
0.772
|
0.646
|
CONCLUSION AND SUGGESTIONS
Figure 9 shows the dashboard, which provided a structured and clear visualization of the impact of the
Random Forest model on the academic performance of the experimental group.
A
summary
card
was
added
showing
an
average
increase
of
45.24%,
evidencing
an
overall
improvement in academic performance after applying personalized learning objects.
Grouped column charts were used to compare the results of the diagnostic test and the final test per
student,
allowing
individual
progress
to
be
observed
and
confirming
that
most
students
showed
significant improvements.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2026, Volumen VII, Número 4 p 2447.
In addition, the scatter plot helped analyze the relationship between the initial score and the percentage
of
improvement,
demonstrating
that
even
those
who
started
with
low
scores
made
remarkable
progress.
Visual dashboard of the conclusions reached in the assessment of the impact of personalized learning
objects,
using
the
Random
Forest
predictive
model,
including
individual
student
comparisons,
the
percentage improvement achieved, and a table with specific
results.
Graphic 7
Learning Object Impact Assessment
Source:
own elaboration.
The application of diagnostic and summative tests, designed to assess basic mathematical skills in
high school students, revealed a notable improvement in their academic performance. After using the
learning objects suggested by the Random Forest model, st
udents obtained better results in the final
test, indicating progress in their understanding of the topics.
This result supports the use of data science techniques, such as the Random
Forest algorithm, to
identify behavior patterns and learning profiles that enable the recommendation of personalized digital
resources.
The predictive model proved to be an effective educational support tool, classifying learning objects
according
to
variables
such
as
frequency
and
duration
of
use,
perceived
impact
on
grades,
and
motivation, which made it possible to accurately predict the
most appropriate resource for each student
(such as videos, interactive exercises, or simulations).
Thanks to this automated personalization, the individual learning needs were addressed, to achieve an
improve of 45.24%in the academic performance. This was validated through the comparison between
the results, of both instruments as shown in the graphic 8
.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2026, Volumen VII, Número 4 p 2448.
Graphic 8
Experimental comparison
Source:
own elaboration.
Furthermore,
statistical
analysis
revealed
that
progress
was
uniform
across
the
different
groups
evaluated, suggesting that the OA recommendation not only benefits a small group of students, but
also has a widespread effect. This discovery underscores the
importance of incorporating predictive
analytics
tools
into
curriculum
development
and
pedagogical
decision
-
making.
Likewise,
it
was
demonstrated that student motivation plays a crucial role, given that those who showed greater interest
obtained more outst
anding results in the final test.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2026, Volumen VII, Número 4 p 2449.
Estu
diant
e 1
Estu
diant
e 2
Estu
diant
e 3
Estu
diant
e 4
Estu
diant
e 5
Estu
diant
e 6
Estu
diant
e 7
Estu
diant
e 8
Estu
diant
e 9
Estu
diant
e 10
Estu
diant
e 11
GRUPO EXPERIMENTAL Puntaje
Diagnóstico
10
12
8
11
9
13
7
10
11
9
17
GRUPO EXPERIMENTAL Puntaje
Final
15
17
13
16
14
18
12
14
15
13
19
GRUPO EXPERIMENTAL Mejora
5
5
5
5
5
5
5
4
4
4
2
0
2
4
6
8
10
12
14
16
18
20
ITEMS
GRUPO EXPERIMENTAL
REFERENCES
Aros, A. S. (2025) Predictive Analysis of the Use of Learning Objects for High Schools in the Teziutlán
Region [Master's Thesis, Instituto Tecnológico Superior de Teziutlán]
De Medio, C., Limongelli, C., Sciarrone, F., & Temperani, M. (2019).
MoodleREC: A recommendation
system for creating
courses using the Moodle
e
-
learning platform.
Computers in Human Behavior.
https://doi.org/10.1016/j.chb.2019.106168
Hadayat Ali, H., Muzammal Hussain, M., Fazil, H., Amjad, F., & Nazir, M. (2024). Navigating the future:
Trends and innovation in Education. Qualitative Research, 345
-
366.
Hernández Osorio, J. del
C. (2023). La Deserción Escolar en la Educación Media Superior. Ciencia
Latina Revista Científica Multidisciplinar, 7(6), 1676–1700. https://doi.org/10.37811/cl_rcm.v7i6.8801
Hotz,
N.
(December
9,
2024).
What
is
CRISP
DM?
Retrieved
from
Data
Science
PM:
https://www.datascience
-
pm.com/crisp
-
dm
-
2/
J. L. García Cué, J. M. Meraz Escobar, & M. A. Jiménez Velázquez, Learning objects as support in the
teaching
-
learning process (p. 87). Mexico: General Directorate of Publications
National Institute of Statistics and Geography. (2025). Interactive tables
-
Generic. Retrieved from INEGI:
https://www.inegi.org.mx/app/tabulados/interactivos/?pxq=Educacion_Educacion_12_987437b7
-
5398
-
4067
-
8b19
-
a1cd2ec36b7b
National Institute of Statistics and Geography. (2023). National Survey on the Availability and Use of
Information
Technologies
in
Households.
Retrieved
from
INEGI:
https://www.inegi.org.mx/programas/endutih/2023/
Global Compact. (January 23, 2025). 17 Development Goals to transform the world. Retrieved from
Global Compact Network Spain:
https://www.pactomundial.org/que
-
puedes
-
hacer
-
tu/ods/
OpenAI. (2024). ChatGPT (Versión GPT-4o) [Modelo de lenguaje grande].
https://chatgpt.com
Orellana-Cordero, M., García-Herrera, D. G., Erazo-Álvarez, J. C., & Narváez-Zurita, C. I. (2020). Objetos
virtuales
interactivos
con
Genial.ly:
Una
experiencia
de
aprendizaje
matemático
en
bachillerato.
CIENCIAMATRIA, 6(3), 309–332.
https://doi.org/10.35381/cm.v6i3.403
Palmera Quintero, L. M., Muñoz Morales, L. A., Guerrero Sumalave, D. F., & Holguín Uribe, Y. (2023).
Objeto
virtual
de
aprendizaje
para
la
enseñanza:
instrumento
de
apoyo
para
la
inclusión
de
la
informática
en
niños
con
discapacidad
visual.
Mundo
Fesc,
13(27),
38–49.
https://doi.org/10.61799/2216-0388.135439
Pulido Ocampo, C. A., Martínez Moreno, P., & Vergara Camacho, J. A. (2016).
Learning objects as a
teaching resource in the Algorithmic educational experience. In J. A. Hernandez Moral, A. Ronquillo
Bolaños, A. B. Celis Domínguez, A. Lagunes Domínguez, A. P. Lorandi Medina, & A. D. Otero Escobar,
Trends and challenges in educationa
l innovation: an open debate.
Mexico: Foundation for International
Higher Education.
Rodríguez,
N.
W.
(2023).
Objetos
virtuales
de
aprendizaje
en
el
marco
de
la
educación
4.0:
una
alternativa pedagógica mediada con las TIC para el logro de aprendizajes significativos. Dialéctica,
1(20).
https://doi.org/10.56219/dialctica.v1i20.1669
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2026, Volumen VII, Número 4 p 2450.
Serna Martínez, R. E., & Alvites Huamaní, C. G. (2021). Plataformas educativas: herramientas digitales
de
mediación
de
aprendizajes
en
educación.
HAMUT’AY,
8(3),
66.
https://doi.org/10.21503/hamu.v8i3.2347
Tovar Vidal, M., Zepeda Cortés, C., & Castillo Zacatelco, H. (2019). Objetos de Aprendizaje Abiertos
basados en ADDIE. Caso de estudio: Colegio de Posgraduados. En J. L. García Cué, R. C. Medina, & M.
A. Jiménez Velázquez (Eds.), Los Objetos de Aprendizaje como apoyo en el proceso de enseñanza-
aprendizaje (pp. 87-105). Colegio de Posgraduados.
Triola F. Mario. (2004). Estadística (Pearson Educación, Ed.; 9a ed.).
Todo el contenido de LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, publicados en este
sitio está disponibles bajo Licencia
Creative Commons
.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2026, Volumen VII, Número 4 p 2451.
APPENDIX
During the course of this research, artificial intelligence (AI) tools were used to assist with tasks such
as preliminary drafting of sections of the document and grammatical correction. The use of AI was
limited
to
auxiliary
functions,
and
all
methodologi
cal
decisions,
interpretations
of
results,
and
conclusions were made by the authors; AI did not intervene in the methodological design or statistical
analysis.
The
article
is
derived
from
the
thesis
“Predictive
Analysis
of
the
Use
of
Learning
Objects
for
High
Schools
in
the
Teziutlán
Region”
presented
by
Sara
Aros
Alberto
to
obtain
a
Master's
degree
in
Computer Systems at the Instituto Tecnológico Superior de Tez
iutlán on September 3, 2025.
LATAM Revista Latinoamericana de Ciencias Sociales y Humanidades, Asunción, Paraguay.
ISSN en línea: 2789-3855, septiembre, 2026, Volumen VII, Número 4 p 2452.