Publication: Analysis and short-term prediction of urban traffic in Madrid under the influence of multitudinary events using machine learning techniques
Loading...
Official URL
Full text at PDC
Publication Date
2018
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Abstract
En este trabajo se busca analizar a partir de los datos de tráfico de la ciudad de Madrid, y de ciertos eventos que suceden en la ciudad, la aparición de situaciones de congestión de tráfico. El objetivo es evaluar si estos eventos tienen influencia negativa en el tráfico urbano y en qué puntos de la ciudad. En la medida de lo posible, se busca también alertar de forma anticipada acerca de esas congestiones.
El anterior análisis se ha organizado en varias partes. Primero se presenta un análisis de la información recabada. En él, se pretende observar cuándo y en qué puntos de las vías urbanas ciertos eventos con un desplazamiento masivo de gente resultan más influyentes negativamente en el desarrollo del transporte. A partir de esta información pasada, se realiza una propuesta para un sistema de aprendizaje y clasificación que sea capaz de utilizarla para predecir estos problemas en situaciones similares con algunas horas de anticipación.
Respecto a los datos recabados, han sido datos históricos de tráfico y eventos en la ciudad. Los datos de tráfico recogen métricas, (ej. la intensidad de vehículos o la carga de la vía en un intervalo de tiempo) en diferentes puntos geográficos. A fin de identificar eventos relevantes, también se han recopilado para las fechas consideradas información sobre grandes eventos de masas (ej. partidos de fútbol o periodos de compras). Estos datos se han preprocesado, incluyendo tareas de homogenización de las representaciones, y limpieza de valores perdidos o incompletos. También se ha cruzado información de forma que fuera coherente con el análisis que se pretendía realizar. Tras ello se han aplicado algoritmos de clasificación y aprendizaje (ej. árboles de decisión o k-vecinos más cercanos) y se ha desarrollado el modelo previamente mencionado.
Finalmente, se han expuesto los resultados y conclusiones obtenidos del proceso. En ellos se exploran los resultados obtenidos por los algoritmos clasificadores , se explica cómo afectan los diferentes eventos a los puntos geográficos en términos de las mediciones que se observan en ellos, y se sugieren formas de mejorar los modelos en futuros trabajos.
Este trabajo ha sido realizado con el apoyo de los siguientes proyectos: “Diseño Colaborativo para la promoción del bienestar en Ciudades Inteligentes Inclusivas' (TIN2017- 88327-R) financiado por el Ministerio para la Economía y la Competitividad de España; y MOSIAGIL-CM (S2013/ICE-3019) co-financiado por la Comunidad de Madrid y los fondos estructurales de la Unión Europea FSE, y FEDER.
This work aims at analyzing the appearance of traffic congestions in the city of Madrid. The analysis uses data on traffic and certain events in the city. In this context, the goal is twofold. First, to study which of these events have a real influence in the generation of these congestions and where in the city. Second, being able to alert with some anticipation about these situations. The previous analysis has been organized in several parts. First, an analysis of the considered information is introduced. Its goal is to observe when and in which points of the urban roads certain events that imply a massive displacement of people have a more negative impact on traffic. Then, from this information, a proposal of a learning and classification system able to use that information to predict these problems in similar situations is made. Regarding data, the work gathered historic data on traffic and events. Traffic data include metrics (e.g. intensity of vehicles or road load in a time interval) in different geographical points. Information on events includes football matches of the main city teams and the Christmas shopping period. These data have been preprocessed, including tasks of homogenization of representations, and cleaning of lost or incomplete values. Information has also been merged to make it coherent with the intended analysis techniques. After that, the work applied several classification algorithms (e.g. decision trees and k-nearest neighbors) and developed the prediction model previously mentioned. Finally, results and conclusions have been discussed. This explores the results the classifier algorithms obtained, explains how the different events affect to the geographic points in terms of the observed traffic data, and suggests ways to improve them. This work was supported by the following projects: “Collaborative Design for the Promotion of Well-Being in Inclusive Smart Cities'” (TIN2017-88327-R) funded by the Spanish Ministry for Economy and Competitiveness; and MOSI-AGIL-CM (S2013/ICE-3019) co-funded by Madrid Government, EU Structural Funds FSE, and FEDER.
This work aims at analyzing the appearance of traffic congestions in the city of Madrid. The analysis uses data on traffic and certain events in the city. In this context, the goal is twofold. First, to study which of these events have a real influence in the generation of these congestions and where in the city. Second, being able to alert with some anticipation about these situations. The previous analysis has been organized in several parts. First, an analysis of the considered information is introduced. Its goal is to observe when and in which points of the urban roads certain events that imply a massive displacement of people have a more negative impact on traffic. Then, from this information, a proposal of a learning and classification system able to use that information to predict these problems in similar situations is made. Regarding data, the work gathered historic data on traffic and events. Traffic data include metrics (e.g. intensity of vehicles or road load in a time interval) in different geographical points. Information on events includes football matches of the main city teams and the Christmas shopping period. These data have been preprocessed, including tasks of homogenization of representations, and cleaning of lost or incomplete values. Information has also been merged to make it coherent with the intended analysis techniques. After that, the work applied several classification algorithms (e.g. decision trees and k-nearest neighbors) and developed the prediction model previously mentioned. Finally, results and conclusions have been discussed. This explores the results the classifier algorithms obtained, explains how the different events affect to the geographic points in terms of the observed traffic data, and suggests ways to improve them. This work was supported by the following projects: “Collaborative Design for the Promotion of Well-Being in Inclusive Smart Cities'” (TIN2017-88327-R) funded by the Spanish Ministry for Economy and Competitiveness; and MOSI-AGIL-CM (S2013/ICE-3019) co-funded by Madrid Government, EU Structural Funds FSE, and FEDER.
Description
Facultad de Informática, Departamento de Ingeniería del Software e Inteligencia Artificial, curso 2017-2018