Analyzing Residential Travel Mode Choice Based on Principal Component Analysis

Abstract

A disaggregate model has a wide application in analyzing residential travel mode choice due to its high accuracy of modeling and good interpretability. However, the multicollinearity among explanatory variables, such as family income and desire of buying private cars, seriously influences modeling results as well as accuracy of forecasting. In order to overcome this deficiency, the explanatory variables are first classified according to their correlation. Then those with high dependence are chosen to perform principal component analysis, which is used to develop the multinomial Logit (MNL) model. Survey data of daily travel behaviors of residents living along rail transit in Nanjing City in 2009 are used to validate the effectiveness of the proposed model and also to compare to existing MNL models. The results show that the proposed MNL model based on principal component analysis can efficiently reduce the influence of the multicollinearity among explanatory variables and improve the accuracy of modeling and Macfadden coefficients.