PREDICCIÓN DE LA UBICACIÓN FINAL DE UN AUTOMÓVIL POR MEDIO DE MODELOS DE APRENDIZAJE COMPUTACIONAL

ZUBIA HERNANDEZ, NORMANDO ALI (2016) PREDICCIÓN DE LA UBICACIÓN FINAL DE UN AUTOMÓVIL POR MEDIO DE MODELOS DE APRENDIZAJE COMPUTACIONAL. Maestría thesis, UNIVERSIDAD AUTONOMA DE CHIHUAHUA.

[img]
Vista Previa
Text
Tesis_Normando_Zubia.pdf

Download (4Mb) | Vista Previa

Resumen

Between August 2014 and September 2016, several machine learning algorithms were implemented to tackle the vehicles destination prediction problem. That problem was caused by the electronic dispatching systems updating in taxi services, where it’s easy to know where the taxi has been but not necessarily where it’s going. To solve this problem we use a dataset with 1.5 million trips of 442 taxi vehicles in the city of Porto, Portugal. The dataset was preprocessed since several inconsistencies were found into the trajectories, a neural network was trained with that information, and we got a mean error of 1.8 km between the real destinations and the predicted ones. Results are promising, but the analysis of them suggests that the trajectories similitudes caused by transportation infrastructure (ex. highways) represent a barrier for the correct prediction of final destinations. To tackle this issue, an hybrid approach is proposed in this work, combining the individual patters of each user with the general ones obtained by the neural network to improve the model accuracy.

Tipo de Documento: Tesis (Maestría)
Palabras Clave: Keywords: neural network, trajectory mining, machine learning, patterns combination.
Clasificación temática: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Usuario Remitente: Admin Administrador del Respositorio
Depositado: 06 Nov 2017 16:37
Ultima Modificación: 06 Nov 2017 16:37
URI: http://repositorio.uach.mx/id/eprint/130

Actions (login required)

Ver Objeto Ver Objeto

Universidad Autonoma de Chihuahua