Comparison of Different Mode Choice Models for Work Trips using Data Mining Process

Abstract

Objectives: An attempt made in this paper to compare the prediction ability of different mode choice behaviour models of commuters’ work trip. Methods/Statistical Analysis: The personal characteristics like age, income, family size, occupation of commuters and the trip characteristics such as trip time and trip cost were selected as independent variables. Four types of alternative modes were considered for developing the models such as private car, two wheelers, Shared Auto and bus. The behaviour predicting ability of three different models namely boost tree, MNL and SVM model compared using data mining process in Statistica software. Findings: We found that the boost tree model is superior among all. Applications/ Improvements: Next step would be to find and to put the probabilities of each mode for each zone. Further to assign the route using all or nothing assignment method or capacity restraint assignment method.