Desarrollo de modelos para la identificación de actividades simultáneas en encuestas de uso de tiempo.
Loading...
Date
2025
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad de Concepción
Abstract
El tiempo es un recurso valioso, más que simples horas y minutos, es una medida que ayuda a comprender a las personas y su comportamiento a nivel socioeconómico y cultural, conociendo la forma en que lo utilizan. Las encuestas de uso de tiempo son herramientas vitales para esta medición. Aun así, no se libran de problemas como la dificultad para captar actividades simultáneas, las cuales brindan una representación más realista del consumo del tiempo diario. Este estudio se basó principalmente en la encuesta ENUT 2023, la cual incluyó un módulo enfocado en captar la simultaneidad, por lo cual se evaluó el impacto del módulo en el ajuste de reportes que exceden el presupuesto horario. También se cuantificó la influencia de la simultaneidad de actividades en el valor económico percibido del tiempo. Y finalmente se identificaron patrones que logren predecir la presencia de simultaneidad en encuestas que carezcan de un módulo específico, aplicando el análisis a la ENUT 2015.
Metodológicamente, se realizó la limpieza de datos y análisis de distribución de la ENUT 2023, evaluando la cantidad de sobre reportes incluyendo la simultaneidad y omitiendo. Para las predicciones de patrones que identifican la simultaneidad se emplearon modelos de aprendizaje automático del tipo Random Forest, entrenados con la encuesta de 2023 y siendo aplicado a la edición de 2015. La valoración del tiempo se estimó comparando el Valor del Tiempo de Ocio (VoL) y Valor del Tiempo Asignado al Trabajo (VTAW) con y sin la inclusión de simultaneidad, utilizando un modelo microeconómico basado en la teoría de Utilidad de Jara-Diaz.
Los resultados indican que es posible predecir la simultaneidad, pero con baja precisión, atribuido a: la complejidad del comportamiento humano, las limitaciones de los datos disponibles y modelos utilizados. No obstante, el impacto que obtuvo la simultaneidad en la corrección de sobrereportes resultó sin un impacto significativo mostrando que, el problema no radica solo en la simultaneidad. Similarmente, la influencia de esta en el valor percibido del tiempo a pesar de existir resultó ser poco significativa, que se asocia a la simplicidad del modelo y la necesidad de capturar relaciones más complejas. En conclusión, este estudio no descarta el potencial de la simultaneidad para una comprensión más compleja del tiempo. Sin embargo, subraya la necesidad de mejorar la calidad y disponibilidad de datos sobre actividades combinadas en futuras ediciones de las encuestas.
Time is a valuable resource; more than just hours and minutes, it is a measure that helps us understand people and their behavior at a socioeconomic and cultural level, understanding how they use it. Time-use surveys are vital tools for this measurement. However, they are not free from problems such as the difficulty in capturing simultaneous activities, which provide a more realistic representation of daily time consumption. This study was primarily based on the ENUT 2023 survey, which included a module focused on capturing concurrency. Therefore, the impact of the module on adjusting reports that exceed the hourly budget was evaluated. The influence of concurrent activities on the perceived economic value of time was also quantified. Finally, patterns were identified that could predict the presence of simultaneity in surveys lacking a specific module, applying the analysis to the 2015 ENUT. Methodologically, data cleaning and distribution analysis were performed for the 2023 ENUT, assessing the number of overreports including simultaneity and omitting it. Random Forest machine learning models were used to predict patterns that identify simultaneity, trained with the 2023 survey and applied to the 2015 edition. Time valuation was estimated by comparing the Value of Leisure Time (VoL) and (VTAW) with and without the inclusion of simultaneity, using a microeconomic model based on the Jara-Diaz Utility Theory. The results indicate that simultaneity can be predicted, but with low accuracy, attributed to the complexity of human behavior, the limitations of the available data, and the models used. However, the impact of simultaneity on correcting overreporting was not significant, demonstrating that the problem is not solely rooted in simultaneity. Similarly, its influence on the perceived value of time, despite its existence, was found to be insignificant, which is associated with the simplicity of the model and the need to capture more complex relationships. In conclusion, this study does not rule out the potential of simultaneity for a more complex understanding of time. However, it underscores the need to improve the quality and availability of data on combined activities in future editions of the surveys.
Time is a valuable resource; more than just hours and minutes, it is a measure that helps us understand people and their behavior at a socioeconomic and cultural level, understanding how they use it. Time-use surveys are vital tools for this measurement. However, they are not free from problems such as the difficulty in capturing simultaneous activities, which provide a more realistic representation of daily time consumption. This study was primarily based on the ENUT 2023 survey, which included a module focused on capturing concurrency. Therefore, the impact of the module on adjusting reports that exceed the hourly budget was evaluated. The influence of concurrent activities on the perceived economic value of time was also quantified. Finally, patterns were identified that could predict the presence of simultaneity in surveys lacking a specific module, applying the analysis to the 2015 ENUT. Methodologically, data cleaning and distribution analysis were performed for the 2023 ENUT, assessing the number of overreports including simultaneity and omitting it. Random Forest machine learning models were used to predict patterns that identify simultaneity, trained with the 2023 survey and applied to the 2015 edition. Time valuation was estimated by comparing the Value of Leisure Time (VoL) and (VTAW) with and without the inclusion of simultaneity, using a microeconomic model based on the Jara-Diaz Utility Theory. The results indicate that simultaneity can be predicted, but with low accuracy, attributed to the complexity of human behavior, the limitations of the available data, and the models used. However, the impact of simultaneity on correcting overreporting was not significant, demonstrating that the problem is not solely rooted in simultaneity. Similarly, its influence on the perceived value of time, despite its existence, was found to be insignificant, which is associated with the simplicity of the model and the need to capture more complex relationships. In conclusion, this study does not rule out the potential of simultaneity for a more complex understanding of time. However, it underscores the need to improve the quality and availability of data on combined activities in future editions of the surveys.
Description
Tesis presentada para optar al título de Ingeniero/a Civil Industrial.
Keywords
Cronobiología, Encuestas Chile, Cronometría