Big Traffic Card Data Analysis Using Spatial Seemingly Unrelated Regression Model: A Case Study for Gyeonggi-do Province

Seemingly unrelated regressions are a type of multi-equational econometric formulation. The SUR framework assures increased efficiency by estimating the system jointly rather than by processing each equation separately. Also, in recent decades we have seen an increased interest in the use of seemingly unrelated regressions models (SUR) in spatial context, in which incorporate spatial effects and correlated error terms across equations. In this paper, we examine big transportation card users of 736,873 observations about individual bus card tagging information, from July 1 to July 4, 2018, Gyeonggi-do province, South Korea. At first, we obtain Origin-Destination matrix and kernel density plot of hour, and so on, during the phase of EDA. And we consider two dependent variables : “Within” and “Exit”. “Within” is defined by the number of intra-provincial trips and “Exit” is defined by the inter-provincial trips, in province “i” and day “t”. We analyze the trip data using SUR-spatially independent model (SUR-SIM), SUR-spatial lags in the Xs model (SUR-SLX), SUR-spatial lag model (SUR-SLM), and SUR-spatial error model (SUR-SEM), with various weighting matrices and many independent variables.