Maestría Estadística Aplicada

URI permanente para esta colecciónhttp://hdl.handle.net/11634/13159

Examinar

Mostrando 1 - 20 de 62

Predicción espacio-temporal y análisis de tendencias del PM10 y PM2.5 en zonas de nuevas edificaciones en bogotá en el periodo de 2012-2021
(Universidad Santo Tomás, 2026-04-29) Jaimes Cuberos, Ederson; Pineda Rios, Wilmer Darío; Universidad Santo Tomas; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0001454199; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0000800147; https://scholar.google.com/citations?user=4-t7xVcAAAAJ&hl=es&oi=ao; https://orcid.org/0000-0001-7774-951X; https://orcid.org/0000-0002-4407-7646
This study aimed to estimate the concentrations of particulate matter PM10 and PM2.5 in areas with new building developments in Bogotá and to analyze their spatio-temporal trends during the period 2012-2021. The expansion of residential construction in the city responds to population growth and increasing housing demand across several localities. Data processing, integration, and cleaning were carried out using R software (version 4.5.1, 2025-06-13). The "spacetime" and "gstat" packages were employed to construct the empirical spatio-temporal variogram, estimate the spatio-temporal covariance structure, fi t valid theoretical variogram models, and perform spatio-temporal interpolation through kriging. Prior to these procedures, missing data were imputed using the Kalman filter through the "imputeTS" package, considering only monitoring stations with at least 70% data availability for PM10 and PM2.5; stations not meeting this criterion were excluded from the analysis. Subsequently, the spatio-temporal variogram was constructed, and concentrations of PM10 and PM2.5 were predicted at unobserved locations and times associated with new building sites, using the ordinary kriging method with a Best Linear Unbiased Predictor (BLUP). This approach enabled the estimation of the average particulate matter concentration across the full spatio-temporal domain of interest, providing a detailed characterization of pollutant levels in urban areas undergoing development. To evaluate long-term tendencies in the estimated concentrations, a Mann-Kendall (package "trend") non-parametric trend analysis was applied to the interpolated time series for PM 10 and PM 2.5 at 101 locations associated with new developments. The results revealed signi cant decreasing trends for both pollutants during 2012-2022. Overall, the methodological framework demonstrates that spatio-temporal geostatistical modeling is a robust approach for estimating air pollutant concentrations in areas lacking direct monitoring, and for identifying long-term trends relevant to urban environmental planning.
Predicción de la producción de rosas mediante aprendizaje automático: Una comparación entre modelos fenológicos y autorregresivos
(Universidad Santo Tomás, 2026-04-14) Caicedo Arroyave, Lina Constanza; Pineda Ríos, Wilmer Darío; Zainea Maya , Carlos Isaac; Universidad Santo Tomás; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0001454199; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0001449182; https://scholar.google.com/citations?user=4-t7xVcAAAAJ&hl=es&oi=ao; https://scholar.google.com/citations?user=M4uOkgMAAAAJ&hl=es&oi=ao; https://orcid.org/0000-0001-7774-951X; https://orcid.org/0000-0002-6459-2397
This research compares the effectiveness of two modeling approaches for predicting rose production: one based on traditional phenological counts and another purely autoregressive, using classical Machine Learning (ML) algorithms as a low-cost alternative. Due to the limited availability of historical data, synthetic datasets were generated to preserve the seasonal and cyclical patterns of Colombia’s floriculture sector. MLP, LSTM, and XGBoost models were evaluated under a reproducible experimental design, applying cross-validation and standard error metrics (MSE, R²). Results indicate that the autoregressive XGBoost model achieved the best performance (R²=0.82), outperforming models based on phenological information (R²=0.809). These findings demonstrate that production history provides a stronger predictive signal than manual field counts, reducing dependence on subjective and labor-intensive procedures. The study offers a replicable predictive framework that enhances production planning and strengthens the competitiveness of the floriculture industry through efficient use of existing data.
Modelo de regresión beta con valores daltantes en las covariables: Una aplicación sobre la tasa de pobreza multidimensional de los municipios de Colombia.
(Universidad Santo Tomás, 2026-01-19) Barajas Pérez, Rolando Javier; Pacheco López, Mario José; Universidad Santo Tomás; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0000775479; https://scholar.google.com/citations?user=a5SEoPgAAAAJ&hl=es&oi=ao; https://orcid.org/0000-0003-4752-703X
In many studies modeling variables of interest, it is common to encounter missing data for the explanatory variables. This presents a challenge in terms of how to properly handle this data, as the very nature of the variables influences the strategies that should be implemented. In this regard, one of the most frequently used strategies for handling missing data is imputation. While there are various ways to apply this process, it is not always done correctly, since in some cases the fact that the chosen statistical method and imputation method are closely linked is not taken into account. Therefore, a methodology was developed for handling missing values in explanatory variables within a beta regression model, allowing the modeling of proportion or ratio variables, such as the poverty rate. This is a topic of great interest for public policy formulation, especially for generating actions and projects focused on improving the quality of life for people in the country.
Predicción del precio de la posición de las exportaciones de flores en Colombia usando modelos de aprendizaje de máquina
(Universidad Santo Tomás, 2025-09-30) Hernandez Soto, Sebastian Jose; Sierra, Javier Mauricio; Universidad Santo Tomas; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0001533421; https://scholar.google.com/citations?user=WPVb1csAAAAJ&hl=es&oi=ao; https://orcid.org/0009-0003-5914-4156
Exports in emerging economies, such as Colombia, play a vital role in various processes that drive economic dynamics. For instance, 2020 was recorded as a year of recession, where the global economy contracted by 4.4 %. Advanced economies experienced a decline of 5.8 %, while emerging economies in Latin America faced a sharper contraction of 8.1 %. In Colombia, the GDP fell by 7 % due to the pandemic, representing a 3.8 % decrease compared to the GDP of 2019. The restrictions imposed during the pandemic had a significant impact on the flower export market, which, according to Ministerio de Agricultura y Desarrollo Rural de Colombia (2024), contributes to the national economy by generating 130,000 rural jobs. Considering these fluctuations in international markets, it is essential to develop predictive statistical models to understand how key markets for Colombia will behave. Therefore, this research aims to develop predictive models using machine learning techniques, such as Neural Networks LSTM, Random Forests, and XGBoost, as well as traditional statistical models like time series analysis, to forecast export prices by tariff position in Colombia’s flower market.
Estimación del Porcentaje de Afiliación a Seguridad Social en Salud en Bogotá, Utilizando Modelos de Estimación en Áreas Pequeñas con Información Auxiliar Medida con Error
(Universidad Santo Tomás, 2025-09-23) Martínez Salazar, Paola Andrea; Ortiz Rico, Andrés Felipe; Universidad Santo Tomás; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0000650579; https://scholar.google.com/citations?user=xDebiZgAAAAJ&hl=es&oi=ao
Small area estimation is a methodology used to produce reliable estimates of a characteristic of interest by incorporating auxiliary information through a model, thereby improving the accuracy of direct estimates. Since in many cases this auxiliary information contains measurement error, it becomes necessary to account for this component in the proposed model. The application is carried out using the Bogotá Multipurpose Survey – EM (2021), in order to obtain relevant and reliable information on the percentage of health social security affiliation across different life stages, which makes it possible to identify population groups that may have specific medical care needs. Estimating the percentage of affiliation to health social security is crucial for informed decision-making and the design of public policies in Colombia, especially in the context of the Sustainable Development Goals (SDGs), such as promoting health and well-being, reducing inequalities, and ensuring universal health coverage. The proposed methodology allows for comparisons between direct estimates and those generated by the model, to determine which are the most accurate and useful. This not only improves the quality of the available data but also strengthens the District’s capacity to respond to the needs of its population more effectively and equitably.
Análisis de la Correlación Espacial de las Precipitaciones en Bogotá, Mediante la Prueba I de Moran para Datos Funcionales
(Universidad Santo Tomás, 2025-09-19) Miranda Rivera, Pedro Nel; Pineda Ríos, Wilmer Darío; Universidad Santo Tomás; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0001454199; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0001752754; https://scholar.google.com/citations?user=4-t7xVcAAAAJ&hl=es&oi=ao; https://orcid.org/0000-0001-7774-951X
This degree work is a quantitative and descriptive project, based on data on precipitation data recorded at the 91 pluviometric stations operated by IDEAM (Institute of Hydrology, Meteorology and Environmental Studies) in the city of Bogotá, with data collected between 2018 and 2022. Upon reviewing the dataset from the 91 stations, it became necessary to reduce the number of stations to 19, as these provided the most consistent data—both in terms of the higher number of recorded observations and the lower incidence of NA ("Not Available") values. By measuring and analyzing the spatial correlation among these selected stations using a modified version of Moran’s I test—adapted for spatial data and applied to the functional data derived from this refined dataset—it was concluded that there is no spatial correlation between the stations when analyzing the precipitation variable. This result was confirmed using five different methods of measuring distances between stations, both across the overall set of 19 stations and within five specific subsets that were examined.
Riesgo Operacional: Análisis Funcional de los Eventos Materializados Desde los Estados Financieros y las Transacciones
(Universidad Santo Tomás, 2025) Montejo Díaz, Omar Fernando; Pineda-Ríos, Wilmer Darío; Beltrán, Óscar; Univeridad Santo Tomás; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0001454199; https://scholar.google.com/citations?user=4-t7xVcAAAAJ&hl=es&oi=ao; https://orcid.org/0000-0001-7774-951X
The objective is to analyse operational risk (OR) expenses in banking institutions, taking into account potential relationships with financial statement variables, such as dividend income and interest-bearing assets, among others. Additionally, the relationship between OR expenses and monetary transactions carried out through channels provided by banks for their clients will be assessed. To carry out this analysis, the functional data analysis (FDA) methodology will be employed, consisting of the following stages: Functional descriptive analysis: this stage will enable a detailed and continuous understanding of the data over time. Functional clustering analysis: banking entities with similar behaviour regarding OR expenses and explanatory variables will be grouped. Indicator construction: functional principal component analysis (FPCA) will be used to create indicators that facilitate OR monitoring. Functional prediction: the functional regression methodology will be applied to forecast the beha viour of OR expenses based on financial and transactional variables of credit institutions.
Comparación de Cinco Modelos de Machine Learning para la Predicción de las Elecciones Presidenciales en Colombia: una Perspectiva con Datos Composicionales
(Universidad Santo Tomás, 2025-04-02) Leal Varón, Paula Andrea; Galeano Ortiz, Germán Andrés; Pineda Ríos, Wilmer Darío; Universidad Santo Tomás; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0001454199; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0001420586; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0001882179; https://scholar.google.com/citations?user=Id9O7TkAAAAJ&hl=es&oi=ao; https://orcid.org/0000-0001-7774-951X; https://orcid.org/0000-0002-4306-1159
In recent years, numerous studies have employed machine learning techniques and compositional data analysis in various fields of study. However, their integration into electoral analysis remains limited. For this reason, this work combines both approaches by applying five machine learning models: random forest, gradient boosting, support vector machines, k-nearest neighbors, and feedforward neural networks, to predict the results of the presidential elections in Colombia at the municipal level, onsidering the data as compositional. Specifically, it forecasts the vote distribution in each municipality along a unidimensional Left-Right ideological spectrum. This approach aims not only to improve prediction accuracy but also to comtribute a significant advancement in methodologies applied to electoral analysis. The models were trained on 70% of the presidential election data from 2002 to 2022 and evaluated on the remaining 30%. The algorithms demonstrated similar performance across transformations of each ideological spectrum, with variability percentages between 56% and 94% in predicting vote proportions, with the feedforward neural networks model using the centered log-ratio transformation achieving the best results.
Patrones de Desempeño Académico: Un Modelo para Educación Básica y Media Haciendo uso de Minería de Datos
(Universidad Santo Tomás, 2024) Blanco Niño, Andrés Eduardo; Pineda Ríos, Wilmer Darío; Universidad Santo Tomás; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0001454199; https://orcid.org/0000-0001-7774-951X
Currently, the use of technological tools to analyze student academic performance has become a fundamental necessity for informed decision-making within the educational system, especially in the context following the COVID-19 crisis. The pandemic revealed the importance of understanding not only academic results, but also students' fundamental skills and the socio-emotional factors that influence their learning. In this sense, the present research focuses on identifying academic performance patterns of students at Colegio Bilingüe José Max León in the area of mathematics, paying special attention to the factors that affect their performance. The analysis of data on academic performance in mathematics will provide valuable information that will allow the department of the area to optimize its action plans, curricular matrices, and pedagogical practices. Likewise, it will be of great use to the section directors and the psychology department, who will be able to form support groups and design prevention strategies that address the specific needs of students.
Relación Entre la Contaminación por Material Particulado (PM10 y PM2,5) y la Tasa de Mortalidad por Enfermedades Respiratorias en Bogotá D.C (2018-2022): un Análisis Considerando Modelos Espacio-temporales
(Universidad Santo Tomás, 2025-01-29) González Rivera, Wilmer Alfonso; Guzmán Murcia, María Isabel; Ortiz Rico, Andrés Felipe; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0000650579; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0001647865; https://scholar.google.com/citations?hl=es&user=xDebiZgAAAAJ; https://orcid.org/0009-0009-8590-5231
Air pollution represents a significant risk to human health, as such externalities directly affect the quality of life of millions of human beings around the world, thus, the scientific community widely supports the link between pollution and respiratory diseases. According to the World Health Organization (WHO), in 2019, ``approximately 37% of premature deaths related to outdoor air pollution were due to ischaemic heart disease and stroke, 18% and 23% to chronic obstructive pulmonary diseases and acute respiratory infections, respectively, and 11% to respiratory tract cancer'' [World Health Organization, 2022]. Therefore, it is relevant to analyse the relationship between air pollution by particulate matter (PM_{10} and PM_{2.5}) and the mortality rate due to acute respiratory diseases in the city of Bogotá D.C. for the period 2018-2022, generating a reference point that allows us to visualise and mitigate this problem. For this purpose, spatial and spatio-temporal models, which combine the spatial and temporal dimension to analyse geographically dispersed data recorded at different points in time, were used, these models allow capturing the spatial and temporal variation of air pollution and mortality, considering spatial and temporal autocorrelation, trends over time, prediction of future values and the interaction between variables in space and time. The findings suggest the need to implement effective measures to control air pollution and reduce mortality from respiratory diseases, highlighting the relevance of considering geographic location and time periods in planning interventions to mitigate the negative impacts of air pollution on the health of Bogotá's inhabitants.
Aproximación espacial sobre la pobreza monetaria en Bogotá: una aplicación desde la estimación en áreas pequeñas
(Universidad Santo Tomás, 2024-12-13) Durán Gil, Carlos Alberto; Téllez Piñérez, Cristian; Ortiz Rico, Andrés Felipe; Universidad Santo Tomás
The incidence of monetary poverty is a fundamental indicator in assessing the socioeconomic conditions of the population, and its monitoring is part of the Agenda for the Sustainable Development Goals (SDG). In response to the growing need for detailed information for monitoring purposes, this work develops a methodology focused on small area estimation (SAE) with the aim of achieving disaggregations and maps of monetary poverty in households at the level of the zonal planning unit (UPZ for its acronym in Spanish) in the city of Bogotá. Based on the microdata derived from the Integrated Household Survey (GEIH for its acronym in Spanish) for the year 2021, and the use of 25 covariates obtained from geospatial data, Fay-Herriot models are carried out in order to obtain the best linear unbiased estimators (EBLUP) along with their robust spatial extensions (RSEBLUP), comparing their precisions through marginal errors. The results obtained show that the covariates used in the models are adequate predictors of monetary poverty, and that the addition of the spatial component to the model, applying robust processes, provides better precision compared to the direct estimates resulting from the survey.
Clasificación De Especies: Un Enfoque Para Mejorar La Precisión En La Identificación De Serpientes
(Universidad Santo Tomás, 2024) Ruiz Sandoval, Diego Ferney; Bru Cordero, Osnamir; Universidad Santo Tomás; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0001425196; https://scholar.google.com/citations?user=jtoOoEIAAAAJ&hl=es&oi=ao; https://orcid.org/0000-0001-9425-9475
This work aims to apply modern data analysis techniques, particularly neural networks, to improve the classification of snake species in Colombia, using information obtained from records and observations shared on social platforms like Facebook, where users document sightings and characteristics of snakes. In addition to the accurate classification of species, the study is expected to identify patterns in the geographical distribution and behaviors of these species. While neural networks offer a significant advantage in terms of accuracy, they have the disadvantage of being ”black box”models, making them difficult to interpret. Despite this limitation, the study builds on evidence from previous research and data from the online community, with the expectation of improving both species identification and understanding of distribution patterns, contributing to the development of more effective conservation strategies and the protection of biodiversity in Colombia.
Automatización en la toma de decisiones para procesos de inversión en la bolsa de valores de Colombia. una aplicación de modelos avanzados de predicción, modelos de optimización de portafolio y teoría de juegos.
(Universidad Santo Tomás, 2024-11-06) Arevalo Herrera, Carlos Mario; Pineda Rios, Wilmer Dario; Universidad Santo Tomás
The stock market is a key driver for the economy. In this paper, we investigate the behavior of stocks in the Colombian Stock Exchange using linear dynamic models, random forests, and neural networks to identify moments of significant growth in their utility. We then apply the Sharpe and Markowitz models to the stocks projected to grow, in order to determine the optimal percentage of investment in each one. The results show that the Sharpe optimization model, applied to the stocks predicted by random forests, offers the highest return (24.4% monthly), followed by the Markowitz model applied to the stocks predicted by random forests (22.7%) and the traditional Sharpe model (9.9%). This underlines the importance of integrating advanced prediction models and considering both historical performance and future projections to achieve more effective investment.
Ecuaciones de Estimación Generalizada para Modelar el Efecto Inward-Outward en Español
(Universidad Santo Tomás, 2024-12-01) Macías Bohórquez, Ricardo; Ortiz Rico, Andres Felipe; Ortiz Rico, Andres Felipe; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0000650579; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0001821332; https://scholar.google.com/citations?user=xDebiZgAAAAJ&hl=es&oi=ao; https://orcid.org/0000-0002-0217-8756
The Inward-Outward effect suggests that oral articulatory movements can influence psychological processes, such as perception and trust, particularly in the context of e-commerce. This study replicated and extended the findings of Silva and Topolinski (2018) to investigate whether consumers perceive providers with names that generate “inward” articulatory movements (into the mouth) as more trustworthy compared to those with “outward” movements (out of the mouth). An experiment was designed involving 156 university undergraduate and graduate students selected by convenience sampling (91 women, 65 men, with a mean age of 26.13 years and a standard deviation of 7.77 years). Participants chose between pairs of Inward-Outward names and then evaluated the trustworthiness of the associated sellers. The results, modeled using Generalized Estimating Equations (GEE), showed that inward names were perceived as significantly more trustworthy than outward names (p < 0.001). However, no direct relationship was found between initial preference and final choices, suggesting that the Inward-Outward effect may function as a heuristic based on articulatory fluency and presentation order, rather than a conscious decision-making process. Ergonomic implications are discussed, and future research is suggested to explore the role of factors such as interface positioning and response time in the perception of trust.
Análisis Estadístico para el Cálculo de la UPC Utilizando Modelos Lineales Generalizados
(Universidad Santo Tomás, 2024) Caballero Otálora, Adriana Marcela; Pineda-Ríos, Wilmer Dario; Universidad Santo Tomás; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0001454199; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0000294543; https://orcid.org/0000-0001-7774-951X; https://orcid.org/0000-0003-3710-923X
The correct estimation of an insurance premium ensures that, when a claim is filed, the insurance company is capable of responding under the agreed conditions. Colombia's healthcare system adheres to this principle, with the Ministry of Health and Social Protection responsible for estimating the amount to be allocated to each EPS (Health Promotion Entity), ensuring coverage of health risks and financial risks, promoting financial sustainability, and providing comprehensive care to affiliates. Therefore, it is necessary to approach models with predictive capabilities that allow for the proper distribution of resources. Generalized linear models will be explored to adequately address the country's needs, modeling frequency and severity independently, with the aim of reducing potential risk selection in the system.
Propuesta de Estimación de Características Sensibles Utilizando Métodos Bayesianos para la Técnica de Conteo de Ítems de Hussain
(Universidad Santo Tomás, 2023-11-17) Moreno Ibagué, Anyi Biviana; Cruz Pérez, Edwin Andrés; Pineda Ríos, Wilmer Dario; Universidad Santo Tomás; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0001525346; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0001454199; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0000101890; https://scholar.google.es/citations?hl=es&user=e6Oad5sAAAAJ; https://scholar.google.es/citations?hl=es&user=5KmOl5oAAAAJ; https://orcid.org/0000-0003-2134-0058; https://orcid.org/0000-0001-7774-951X
This paper presents a Bayesian proposal to improve the Hussain estimator in the Item Counting Technique (ICT), with the objective of eliminating or reducing the proportion of negative estimates that this estimator presents when a particular sensitive characteristic is to be studied. An analytical estimator is proposed using as prior distributions a beta distribution and the uniform distribution, the analysis is carried out via simulation by proposing different scenarios for the number of questions in the questionnaire, sample size and the known proportion of non-sensitive questions, obtaining finally the total elimination of the negative estimates for the proportion of people who have a sensitive characteristic of interest, presenting significant reductions in the estimated coefficient of variation.
Deserción Universitaria en Poblaciones Vulnerables. Estudio de Caso Sobre el Programa "Jóvenes a la U" en Bogotá.
(Universidad Santo Tomás, 2024) Pachón Ariza, Jose Daniel; Pineda Rios, Wilmer; Universidad Santo Tomás; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0001566310; https://orcid.org/0000-0001-7774-951X
Automatización Cartográfica: Integración de Tecnologías para la Exportación Automatizada de Mapas
(Universidad Santo Tomás, 2023-11-13) Morera Robles, Joel Fernando; Ortiz Rico, Andrés Felipe; Universidad Santo Tomás; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0000650579; https://scienti.minciencias.gov.co/cvlac/visualizador/generarCurriculoCv.do?cod_rh=0000037382; https://scholar.google.com/citations?hl=es&user=OuVxcUgAAAAJ

Examinar

Envíos recientes