Analysis of Moving Patterns of Moving Objects with the Proposed Framework

Abstract

This paper proposes an analysis framework which enables us to analyze the moving patterns of moving objects. To show the effectiveness of the framework, we applied the framework to analyze moving patterns of taxis based on the real-life location history data accumulated from the Taxi Telematics system developed in Jeju, Korea. The analysis aims at obtaining value-added information necessary to provide empty taxis with location recommendation services for the efficient operations of taxis. The proposed framework provides the flow chart which would have a quick look around the overall analysis process and help quickly deal with the same or similar analysis, while saving the temporal and economic costs. Data mining tool used in the framework is Enterprise Miner (E-Miner) in SAS which is one of the most widely used statistics packages and can effectively address huge amounts of log data. Especially, we perform the refined analysis by means of doing repeatedly the well-known k-means clustering method under various spatial or temporal conditions. The paper proposes the refined data mining process 1) extracting the interested dataset about meaningful information driven from the previous cluster results, 2) performing again the detailed clustering with the extracted dataset, and 3) finally extracting the value-added information such as the good pick-up spots or 4) returning the feedback. As a result, the spatiotemporal pattern analysis within the each refined clustering method makes it possible to recommend that the empty taxis go to the nearby cluster location with a high pick-up frequency statistically, resulting in the reduction of empty taxi ratio.

This research was supported by the MKE (The Ministry of Knowledge Economy), Korea, under the ITRC (Information Technology Research Center) support program supervised by the IITA (Institute for Information Technology Advancement). (IITA-2009-C1090-0902-0040).