Abstract
In this paper, we apply a factor analysis which is a kind of multivariate statistics to the results of vehicle traffic simulation of Kobe-city, Japan. City scale traffic simulations usually output multivariate results. However, both in the context of data assimilation and in the context of statistical hypothesis testing, such results are often treated as independent variables. Lack of concerning such result correlations results in worse accuracy of the simulation result. Therefore, better statistical comparisons and better statistical analyses between simulation results are strongly required. From the results of factor analysis, it showed that the large number of simulation results were described by 33 factors. Obtained factors which had higher contribution rate were robust against the time and space of origin and destination and against the number of vehicles. Moreover, the higher contribution rate factors can be distinguished from other factors which can be obtained even if simulation settings were completely random. Though existences of such factors are out of range of this study, the knowledge of factor analysis results may help us not only explain the city traffic simply but also improve accuracy of large scale vehicle traffic simulations.
