Modeling and use of accessibility measure in urban settlements: investigation of the effects on mode choice

In this study, accessibility measure in urban settlements and its modeling for the prediction of mode choice is investigated. Activity based potential and utility based types of accessibility measures are considered. Beside the correlation analysis, the measures are also examined for the prediction of the aggregated mode choice by using linear and artificial intelligent approaches. The utility based accessibility is used in discrete choice analysis. The prediction capability of the constructed models including different trip purpose classifications and accessibility versions are also compared.The all land use types of activity based potential measure is found to give positive correlations with all modes for extended trip numbers. It is concluded that this measure can introduce better performance for the investigation of mobility. Utility based accessibility measures are determined as an efficient tool for exhibiting a clear difference between aggregated statistics of public transport and automobile modes. Besides, artificial neural network models are found to give considerable prediction capability with reasonable generalization especially for automobile choice.Logit models show that the utility based accessibility having inner zone activities and rational land use has the highest prediction capability for the trips excluding work and school purposes. The logit models including only accessibility measures indicate that discrete choice can be modeled especially for the limited discrete data conditions. As a result, the thesis proves that utility based accessibility is an efficient tool for the prediction of mode choice for the certain types of trip purposes and land use described in the study.