The Spatiotemporal Correlation Analysis of Urban Traffic Characteristics Using an Adaptive Partitioning Method

Abstract

Non-stationary and nonlinear features in the dynamic trajectories of traffic parameters may stop rich and hidden multi-dimensional knowledge from being extracted, contributing a correlation result with false judgments or ignorable significance. As the fundamental improvement for a dynamic congestion analyzing method, the paper decomposes a time series of traffic data into 2 components based on a polynomial approximation method, called the “trend” and “detail” components. In order to simplify the selection of the fitting function to obtain the trend component, the paper proposes an adaptive partitioning method based on the sensitivity search technology. The given examples prove the effectiveness of the proposed methods for a spatiotemporal traffic congestion study.