APPLICATION OF ANN FOR ACCESS MODE CHOICE FOR A MULTIMODAL TRANSPORT SYSTEM

Abstract

Rapid growth of urbanization in India, has invited many transport related problems. The increased road traffic and transport demand has resulted in congestion, accidents, travel delays and environmental pollution crossing acceptable limits. To ease traffic problems in urban areas many government agencies are promoting use of mass transit systems by the commuters. But the level of service provided by these mass transit systems is fully dependent on its accessibility. Multimodal Transportation System (MMTS) is the potential solution for these transportation problems. In this thesis, access behavior of the commuters along Red Line of the Delhi Metro Service has been analyzed according to their socioeconomic and travel characteristics. Acceptable travel distances for different trip purposes are estimated for different access modes available in study area. Multinomial Logit Model (MNL) and Artificial Neural Network (ANN) based access mode choice models are developed for the revealed preference data. Results show that in vehicle travel time, out vehicle travel time and travel cost are factors which are affecting access mode choice decision significantly. Predictive abilities of MNL and ANN have also been compared. Predictive ability of ANN is found superior to MNL model. MNL models and artificial neural networks are also developed for stated preference data of walk and feeder bus. It was found that commuters give more importance to walk facility and crossing facility while choosing walk as an option. In case of feeder bus, commuters attach more importance to the travel time and travel cost. Finally sensitivity analysis is carried out to analyze variation in output with respect to the change in input values.