Publicación: Diseño y mejora de un predictor de tráfico basado en GNN
Fecha
2023-10
Autores
Editor/a
Director/a
Tutor/a
Coordinador/a
Prologuista
Revisor/a
Ilustrador/a
Derechos de acceso
info:eu-repo/semantics/openAccess
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad Nacional de Educación a Distancia (España). Escuela Técnica Superior de Ingeniería Informática
Resumen
En este trabajo se realiza un estudio sobre el efecto de las formas de construcción de grafos y los efectos de las diferentes características disponibles. La previsión de tráfico es un problema cada vez más importante para la gestión y ordenación de las grandes ciudades. Durante los últimos cincuenta años han ido evolucionando los métodos y las técnicas utilizadas para la tarea de realizar previsiones de tráfico. Desde las primera técnicas puramente estadísticas hasta las actuales basadas en el aprendizaje profundo se ha recorrido un largo camino. Las Graph Neural Networks son la última técnica de moda que se ha aplicado a este problema, siendo un campo que está en rápido y continuo desarrollo. Siendo uno de los principales campos de expansión dentro del aprendizaje profundo. En este trabajo hemos realizado un estudio que ha estudiado el efecto de la forma de construcción de los grafos necesarios y la influencia de las diversas características adicionales al tráfico que se pueden añadir a los nodos. Presentando un conjunto de ellas y comparando su influencia entre ellas.
In this work, a study is carried out on the effect of the forms of graph construction and the effects of the different available characteristics. Traffic forecasting is an increasingly important problem for the management and planning of large cities. Over the last fifty years, the methods and techniques used for the task of traffic forecasting have evolved. From the first purely statistical techniques to the current ones based on deep learning, a long way has come. Graph Neural Networks are the latest fashionable technique that has been applied to this problem, being a field that is in rapid and continuous development. Being one of the main fields of expansion within deep learning. In this work we have carried out a study that has studied the effect of the way the necessary graphs are constructed and the influence of the various additional traffic characteristics that can be added to the nodes. Presenting a set of them and comparing their influence among them.
In this work, a study is carried out on the effect of the forms of graph construction and the effects of the different available characteristics. Traffic forecasting is an increasingly important problem for the management and planning of large cities. Over the last fifty years, the methods and techniques used for the task of traffic forecasting have evolved. From the first purely statistical techniques to the current ones based on deep learning, a long way has come. Graph Neural Networks are the latest fashionable technique that has been applied to this problem, being a field that is in rapid and continuous development. Being one of the main fields of expansion within deep learning. In this work we have carried out a study that has studied the effect of the way the necessary graphs are constructed and the influence of the various additional traffic characteristics that can be added to the nodes. Presenting a set of them and comparing their influence among them.
Descripción
Categorías UNESCO
Palabras clave
Tráfico, Madrid, Aprendizaje profundo, GNN, Grafo, python, pytorch, DGL, Traffic, Deep Learning, Graph
Citación
Centro
Facultades y escuelas::E.T.S. de Ingeniería Informática
Departamento
No procede