The Application Of Clustering Analysis And Classification Analysis In Rail Transit System

In recent years, with the rapid development of rail transit systems, large amount of operation data is deposited in the automatic fare collection (in short form AFC) database. It is very difficult to be understand and analyzed for these data and also difficult normally to obtain knowledge from these data due to its characteristic of large amount and complex features of them when the data of rail transit is processed. The traditional methods of querying and statistic hadn't been satisfied for decision-makers of rail transit of these data, they want to discover regulations in more depth and want to be offered with supporting decisions more effectively for the operation work and decisions of rail transit. Thereby, it is necessary at present that data mining and on line analysis process (in short form OLAP) are the primary development tends.Analysis algorithms of clustering and classification are the main research objects in this paper, the basic theory and their applications of the algorithms are described. The theories and their development processes of clustering and classification suitable for the operation data of rail transit were primarily researched, and decision-tree, time series, partitioning clustering method and etc. to solve the large amount knowledge discovery problems are to be applied of AFC operation data of the rail transit.According to the newer expectation maximization(in short form EM) algorithm and C4.5 algorithm at present, and associated with the time serial of the AFC operation data of rail transit, the deposition data of the AFC central database of rail transit is mined for classification and described, from which the intrinsic pattern in the daily operation procession has been discovered. Through correlation analysis, the EM clustering algorithm and decision classification tree algorithm based on time series analysis are proposed. These two algorithms are suitable for analysis of high dimension and large amount data of rail transit AFC operation as well as can be used to solve the problems such as local convergence and efficiency of the traditional mining algorithm. Thereby, the achievement studied in this paper could produce active promoting actions for analysis large amount of data of rail transit AFC operation.Simultaneously, a decision support system (in short form DSS) has been designed and realized found on multi-dimension cubic, data mining and OLAP in this paper by using relative techniques of data warehouse and etc., based on analysis the history deposition data of rail transit AFC operation database. Therefore, the significative tries have been made to provide the rail transit with accurate and high performance decision support.Finally, the data warehouse and multi-dimension cubic were constructed, the model of data mining is established, the report display of result of data mining is completed, the feasibility, maneuverability, extensibility, facility, algorithm correctness, real time and self-adaptability of the system is completed and validated in this paper through combing business intelligence development tools and using of the preparing processes of data extracting, data transformation and etc.The conclusions indicate that the DSS of rail transit AFC designed and realized based on the techniques of data warehouse, OLAP, and data mining has good real signification and practice application value.