EXTRACTION OF SIMILAR BEHAVIOR FROM SMART CARD DATA

Abstract

With the development of ICT, interest in traffic policy planning by utilizing large varieties of accumulated big data has been increasing. In recent years, graph polishing has been proposed as a new methodology for graph clustering. Graph polishing is one of the graph clustering methods. This method can be used to extract patterns that are similar or related to each other by clarifying the cluster structures in the data. The purpose of this study is to reveal travel patterns of railway passengers by applying graph polishing to smart card data that has been introduced in Kagawa Prefecture, Japan. This study uses 9,008,709 data collected during the 15 months from December 1st, 2013 to February 28th, 2015. This data set includes such information as trip histories and types of passengers. This study uses the graph polishing method to cluster the graph of information about individual rides: riding days of the week, riding time zone, passenger type, the origin station and destination station. As a result, this study succeeded in specifying 141 characteristic travel patterns from the graph.