01477635

Component

A great proportion of transit travel time contributed by dwell time for passengers boarding and alighting. Accurate estimation of bus dwell time can help to improve the accuracy of bus travel time prediction that could enhance the efficiency and reliability of public transportation system. This paper assesses nine different Artificial Intelligence (AI) based approaches alongside traditional Multiple Linear Regression (MLR) method to model and estimate bus dwell time based on data collected from Auckland, New Zealand. The AI based methods include five different Artificial Neural Network (ANN), Support Vector Machine (SVM), Gene Expression Programming (GEP), Decision Tree (DT) and Tree Boost (TB). These methods are widely used in engineering as well as other disciplines, while they have not been applied for bus dwell time modelling and estimation. These methods have been also used to address deficiencies in MLR models, such as, dealing with multicollinearity, interactions between explanatory variables and violation of the normal random error assumption between dependent and independent variables. The study results revealed strengths and weaknesses of these methods for bus dwell time modelling and estimation. Among them, DT and GEP performed reasonably well to model bus dwell time and to overcome problems of MLR models.

This paper was sponsored by TRB committee AP050 Bus Transit Systems.

01470560

13-2495

English

16p

2013

Transportation Research Board 92nd Annual Meeting

Digital/other

Figures; Maps; References (34) ; Tables

Artificial intelligence; Boarding and alighting; Bus transit; Dwell time; Estimating; Intelligent transportation systems; Mathematical models; Public transit; Travel time

Passenger Transportation; Public Transportation; I72: Traffic and Transport Planning

Transportation Research Board Annual Meeting 2013 Paper #13-2495

TRIS, TRB, ATRI

Feb 5 2013 12:32PM