Travel Time Prediction Model For Public Transport Buses In Qatar Using Artificial Neural Networks

Abstract

The state of Qatar has experienced rapid population growth over the last few years. This growth of population has caused authorities to promote the use of public transportation, by introducing new public transport systems such as transit buses and metro lines. The existing bus system was introduced in 2004 to the local community in Qatar. Despite the importance of this system, there are limited studies that are done to analyze and identify its characteristics. There is not much analysis of the stop-to-stop travel time or schedule reliability. The objective of this research is to develop a prediction model for transit route travel time. The model can predict the travel time of buses using several independent variables that are different for each transit route. The prediction model can be used as a useful tool to the decision makers and public transport officials, which can be used for planning, system reliability and quality control, and real-time advanced travelers’ information systems. The data was collected for 12 routes over a period of one year (2015-2016) within The Greater City of Doha using Automatic Vehicle Location (AVL) system. Transit travel time data was obtained from Mowasalat records, the sole operator of public transport buses in Qatar. The collected data include travel time data, route information, geometric configurations, land use, and traffic data. After systematic checking of errors in the collected data and elimination of irreverent records, more than 78,004 trips were analyzed using Artificial Neural Networks (ANN) data mining technique. Prediction model, with R2 of 0.95 was developed. The results indicate that the developed model is accurate and reliable in predicting the travel time. The model can be generalized as well to be applied to newly planned routes, or updating the schedules of existing routes.