Research And Implementation Of Process Visualization Data Mining Tool Based On Crisp-dm

Data mining and knowledge discovery have become research hotspots in the field of computer nowadays.How to get useful information and knowledge from the mass of data,and how to mine unknown regulations hided in data are problems to be urgently solved by human.Technologies related to data mining tools have been researched,and a data mining tool that can be applied on automatic fare collection(in short form AFC) system of rail transit has been designed and implemented in this paper,which combines process model of cross industry standard process for data mining(CRISP-DM) and process visualization technology.Primary research work made by author of this paper is as the followings:(1) Technologies related to data mining,process model and visual data mining have been researched theoretically.Data and business of AFC system of rail transit have been analyzed in detail.(2) Three-tier architecture of data mining tool of AFC system of rail transit (client tier,server tier and database tier) as well as four-level structure(data driving interface layer,data processing layer,data mining and visual displaying layer) has been designed and implemented,which can improve performance of large data processing.(3) CRISP-DM methodology has been researched.According to six phases (business understanding,data understanding,data preparation,modeling,evaluation and deployment) as well as four levels(phase,generic task,specialized task and process instance) of CRISP-DM process model,tasks and outputs of each phase have been designed,data mining context has been used for achieving the mapping between general task level and specialized task level,and reuse of process model has been realized.(4) Research and implement of visualization of data mining process have been focused on.Directed graph has been used to represent and store data mining process. For controlling the interaction and delivery of data flow and command flow,all the nodes of the flow chart have been stored in node table and all the link lines of the flow chart have been stored in link line table.The method of obtaining data source of data mining and the designs of task nodes,control nodes,link lines and data flow diagram have been described in detail.(5) The scalability of data mining tool has been researched.The process and result of using the process visualization data mining tool based on CRISP-DM have been showed by a simple example.Practice shows that operating interface of the data mining tool is flexible and friendly;the tool can be used for data mining,analysis and forecasting of AFC system of rail transit;the tool can improve operating management,increase decision-making ability and reduce the costing of operational maintenance.