Highway Research Record

Qualitative Aspects of Mode Choice Decisions

Date of Start:July 2001
Indian Institute of Technology, Madras (R, C)

Scope and Objective
The scope of the present study is to explore mode choice behaviour based on the utility maximization theory considering the conventional and unconventional factors. The prime objective of the study is to model the mode choice behaviour of users with focus on the attitudinal and perception factors in the context of developing countries like India taking the work trips into account. The other objectives are i) to explore the influence of qualitative aspects in mode choice decisions, ii) to suggest a model framework for work trip travel for a metropolitan city in India, and iii) to suggest alternate methods for modelling.

A suitable survey instrument to capture the attitudes and perceptions of travellers regarding the qualitative aspects is developed. Data pertaining to the home to work travel is collected and analysed for travel behaviour. The mode choice models are developed as logit models using random utility theory. The influence of various qualitative variables and combinations and marginal effects are estimated. The strengths of the conventional methods of modelling and alternate tools like Artificial Neural Network (ANN) in model building are evaluated.

Findings and Conclusions
(i). Survey results in Chennai indicate that,
     a. There is a strong concern regarding the comfort of travel, safety, and pollution among the work trip travellers.
     b. There is lesser concern about cost of travel in comparison to perception of walking time, waiting time and travel time.
     c. There is no major difference in people’s perception on various qualitative aspects with respect to type of trip, age and           location of stay.
(ii). The predictive ability of the multinomial models went up significantly with the inclusion of qualitative variables.
(iii). The best ANN model could produce a success rate of as high as 94 % against that of 86% for the corresponding logit       model with qualitative variables, indicating its potential.