Activity pattern generation using machine learning techniques for transport demand models

Activity pattern generation is an essential component in activity-based demand modeling systems, which are fundamental for urban planning and efficient management of transportation systems. Traditionally, this generation has been performed using conventional techniques; however, with the advancement of technology, machine learning techniques have been increasingly used for tasks such as transportation mode choice and traffic flow prediction. This study focuses on the application of machine learning techniques for the generation of activity patterns for transportation demand models. Four classification tasks will be performed: means of transportation, reason for travel, destination and start time of travel. For these tasks, a dataset of the city of Cuenca belonging to Ecuador of approximately 3000 records had to be prepared, filtered and adjusted. Three machine learning models were used: Random Forest, Decision Tree and Artificial Neural Networks, and their results were analyzed to choose the best one for each classification task. The results show that the best model for the four classification tasks was the Random Forest, in conjunction with the Grid Search Cross Validation hyperparameter fitting technique. Finally, to give utility to the best models, a dataset was built for a simulation using the MATSim tool, achieving favorable results.

• 第一著者: Ajala Ramos, Santiago Alexander (author)
• フォーマット: bachelorThesis
• 言語: eng
• 出版事項: 2024
• 主題: Patrones de actividad; Aprendizaje automático; Redes neuronales artificiales; Activity pattern; Machine learning; Artificial neural network
• オンライン･アクセス: http://repositorio.yachaytech.edu.ec/handle/123456789/851