Measuring the Relationship between Influence Factor and Urban Rail Transit Passenger Flow: Correlation or Causality?

Abstract

The analysis of factors influencing urban rail transit (URT) passenger flow is often a precondition for establishing prediction models. Much past analysis has focused on correlation analysis and does not draw on research on causal mechanisms. In this paper, a novel causal inference method based on transfer information entropy (TIE) is proposed to determine the causality between influence factor and URT passenger flow. As a comparison, the matrix correlation coefficient (RV2 coefficient) is used to analyze the correlation. Taking the URT system in Xi’an, Shaanxi, China, as an example, the factors that may affect passenger flow are introduced and the causality and correlation are calculated. Compared with correlation analysis, the causal inference method can be used to derive the interactive relationship between influencing factors and passenger flow. At the same time, the causal inference method has greater adaptability to the type of passenger flow and the scope of influence. The result can be used for the theoretical support of transit-oriented development (TOD), and can also provide a reference for road traffic planning.