| dc.contributor.author | Vega-Arteaga, Ulises S. | |
| dc.date.accessioned | 2016-10-21T15:20:33Z | |
| dc.date.accessioned | 2023-03-10T17:25:09Z | |
| dc.date.available | 2016-10-21T15:20:33Z | |
| dc.date.available | 2023-03-10T17:25:09Z | |
| dc.date.issued | 2016-10-19 | |
| dc.identifier.citation | Vega Arteaga, U. S. (2016). Implementation and development of traffic speed and flow prediction through Artificial Neural Networks. Trabajo de obtención de grado, Maestría en Sistemas Computacionales. Tlaquepaque, Jalisco: ITESO. | es |
| dc.identifier.uri | https://hdl.handle.net/20.500.12032/71083 | |
| dc.description | In this work we introduce the most recent techniques to predict traffic flow and speed. This work is composed of the following sections: an introduction, a state of the art, and conclusions sections. In the introduction section we see the importance of being able to predict the traffic conditions for speed and flow; we set our hypothesis that is that using the Levenberg-Marquardt training algorithm we’ll be able to find a global minimum for the problem of predicting traffic conditions; we also specify our general objective that is developing efficient algorithms for the traffic prediction for its variables speed and flow; as well as establishing that our scientifical novelty is using the Levenberg-Marquardt as Artificial Neural Network training algorithm. In the state of the art section we present the summarized contents of the most outstanding research papers about traffic prediction. We continue with the presentation of an approach that uses two statistical algorithms for traffic prediction. This information cover spatial impact, and accidents, and construction events. Additionally, it compares the results of outstanding research done with Artificial Neural Networks, and tatistical Methods, to their own statistical method that consists of two statistical methods embedded into only one by using a threshold used to determine which statistical method should be used and when, depending on the road conditions. We also present the results of traffic speed prediction by data mining techniques and a comparative to Artificial Neural Networks. Additionally, we introduce a section that analyze the problem of traffic prediction but only with Artificial Neural Networks done by the company SIEMENS in Germany. We introduce the Los Angeles Department of Transportation (LA DOT) infrastructure used to obtain the speed and flow measurements as well as the set of programs develop by the author of this thesis to retrieve the information from LA DOT, to process this information and calculate the prediction of flow and speed for a specific sensor. We continue with the problem of traffic prediction. In the process of predicting traffic flow, and speed we made our first approach using a Nonlinear Autoregressive Neural Network with External Input in MATLAB and we obtained promising results. However, this Artificial Neural Network does not have the ability to predict multiple outputs, and we transformed it to a Feedforward Neural Network also from MATLAB. The results obtained are impressive because they reduce dramatically the traffic prediction errors. In order to validate our results with the ones in the international research community we use the data from 95 days which is equivalent to 3 months, which is the commonly reported amount of time studied in this kind of problems. We also present an optimization process for the Feedforward Neural Network for both problems speed and flow prediction. Finally, we present a numerical sensibility analysis in order to determine how robust is our Artificial Neural Network. To close this thesis we present our conclusions as a success of using the Levenberg-Marquardt algorithm to train an Artificial Neural Network for the problem of traffic prediction, and we set up the possibilities of exploring the recently found result by using the learning techniques of deep learning. | es |
| dc.description.sponsorship | Consejo Nacional de Ciencia y Tecnología | es |
| dc.description.sponsorship | Continental Guadalajara Services S.A. de C.V | es |
| dc.language.iso | eng | es |
| dc.publisher | ITESO | es |
| dc.rights.uri | http://quijote.biblio.iteso.mx/licencias/CC-BY-NC-2.5-MX.pdf | es |
| dc.subject | Artificial Neural Networks | es |
| dc.subject | Traffic prediction | es |
| dc.title | Implementation and development of traffic speed and flow prediction through Artificial Neural Networks | es |
| dc.type | info:eu-repo/semantics/masterThesis | es |
