Thesis Construcción y análisis de indicadores para la caracterización de patrones de uso a partir de registros de actividad digital
No Thumbnail Available
Date
2026-10
Authors
Journal Title
Journal ISSN
Volume Title
Program
Ingeniería Civil Matemática
Departament
Campus
Campus Santiago San Joaquín
Abstract
Esta memoria analiza registros de actividad generados por AutoCAD, Revit y TimeDoctor con el objetivo de construir variables que permitan describir y comparar distintos patrones de uso entre usuarios. A partir de estos registros se definieron cinco variables: un perfil general de actividad, concentración horaria, diversidad documental, concentración documental y fragmentación temporal. Para identificar perfiles se compararon K-Means y Gaussian Mixture Models (GMM), seleccionándose GMM por entregar grupos más adecuados para su interpretación. Luego se estudió el comportamiento de los indicadores y su relación con los perfiles mediante análisis gráficos y pruebas estadísticas. Los resultados muestran que la diversidad y la concentración documental son los indicadores que presentan la relación más consistente con los perfiles generales. La concentración horaria y la fragmentación temporal también permiten describir diferencias entre usuarios, aunque su relación con los perfiles es menos uniforme entre plataformas. En conjunto, el trabajo muestra que es posible transformar registros operacionales en variables interpretables a nivel de usuario, permitiendo identificar diferencias, relacionar distintas dimensiones de la actividad y caracterizar patrones de uso.
This thesis analyzes activity records generated by AutoCAD, Revit and TimeDoctor in order to build variables that allow different user behavior patterns to be described and compared. From these records, five variables were defined: a general activity profile, hourly concentration, document diversity, document concentration and temporal fragmentation. To identify activity profiles, K-Means and Gaussian Mixture Models (GMM) were compared, with GMM being selected because it produced groups that were more suitable for interpretation. The behavior of the indicators and their relationship with the identified profiles were then analyzed using graphical analysis and statistical tests. The results show that document diversity and document concentration are the indicators with the most consistent relationship with the general activity profiles. Hourly concentration and temporal fragmentation also describe differences among users, although their relationship with the profiles is less consistent across platforms. Overall, the study shows that operational records can be transformed into interpretable user-level variables, making it possible to identify differences, relate different dimensions of activity and characterize usage patterns.
This thesis analyzes activity records generated by AutoCAD, Revit and TimeDoctor in order to build variables that allow different user behavior patterns to be described and compared. From these records, five variables were defined: a general activity profile, hourly concentration, document diversity, document concentration and temporal fragmentation. To identify activity profiles, K-Means and Gaussian Mixture Models (GMM) were compared, with GMM being selected because it produced groups that were more suitable for interpretation. The behavior of the indicators and their relationship with the identified profiles were then analyzed using graphical analysis and statistical tests. The results show that document diversity and document concentration are the indicators with the most consistent relationship with the general activity profiles. Hourly concentration and temporal fragmentation also describe differences among users, although their relationship with the profiles is less consistent across platforms. Overall, the study shows that operational records can be transformed into interpretable user-level variables, making it possible to identify differences, relate different dimensions of activity and characterize usage patterns.
Description
Keywords
Registros de actividad digital, Agrupamiento no supervisado, Patrones de uso, Análisis de indicadores, Modelo de mezcla gaussiana (GMM)
