Modelling and forecasting bus passenger occupation using data-driven methods
Visualitza/Obre
Estadístiques de LA Referencia / Recolecta
Inclou dades d'ús des de 2022
Cita com:
hdl:2117/348039
Tipus de documentProjecte Final de Màster Oficial
Data2021-04-26
Condicions d'accésAccés obert
Tots els drets reservats. Aquesta obra està protegida pels drets de propietat intel·lectual i
industrial corresponents. Sense perjudici de les exempcions legals existents, queda prohibida la seva
reproducció, distribució, comunicació pública o transformació sense l'autorització del titular dels drets
Abstract
Estimating and forecasting passenger demand is one of the major applications for Intelligent Transport Systems (ITS) data sources. Forecasting public bus transport occupations can be useful in order to improve user comfort levels, but also to help with operational decision making to better supply the demand. In the COVID-19 context, informing users with enough anticipation about future occupations can help to avoid crowds and recover the trust in public transportation services. Depending on the considered time horizon to predict, passenger flow estimation can be divided into long-term if predictions aim to forecast average passenger demand for the coming days, weeks, or even longer; and short-term if the predictive period is less than 30 minutes and recent or real-time information is available. This master thesis proposes long-term and short-term occupation forecasting models to estimate bus passenger flow at stop level. Both long- and short-term forecasting approaches are developed using deep learning models such as Deep Neural Networks (DNN) and Long-Short Term Memory (LSTM) networks, and trained with different transportation data sources. The implemented models are intended to be actively used in several real public transport networks from different bus operators. Thus, this work presents training and prediction pipelines that are in charge of pre-processing raw data, building and training the models, and performing predictions on demand. This pipeline system is designed to efficiently and quickly deploy the forecasting models to different cities based on bus operators standardized data sources. Finally, regarding the implantation of these models into real-life industrial settings, explainability techniques are explored to help identify which features are critical to solve occupation prediction and better understand the inner insights of the models.
TitulacióMÀSTER UNIVERSITARI EN INTEL·LIGÈNCIA ARTIFICIAL (Pla 2017)
Col·leccions
Fitxers | Descripció | Mida | Format | Visualitza |
---|---|---|---|---|
155949.pdf | 26,22Mb | Visualitza/Obre |