Research Of Frequent Pattern Mining System On Data Stream

As the development of information technology, the new data streaming models appear in data mining field, it makes a new challenge in the fields. Because of the dynamic characteristics of data streaming, making the existing data mining technology in static data cannot reach effectively with infinite scale for the data, so the information of data streams mining gradually becomes the domestic and international researchers' concerns. The research for data stream mining can be applied in a broad range of living environment, such as telecom industry, large supermarket chain sales industry, multisensor network field and network monitoring domain has its existing application significance. With such a large scale application prospect, believe that data stream mining's technology will be developed rapidly. Based on introducing concepts in data stream mining and algorithms in data mining at the same time, this paper's research focus on the frequent itemset mining problems in data stream, and put forward a frequent parttern mining system on data stream through a summary model based on CAN-tree. Through this progress,The paper represents an improving methods of constructing the summary model and a suitable frequent pattern mining algorithm. Finally,it introduce multiple tests results and analysis of the results. This paper mainly involves the following several aspects of content:1, Introducing data stream mining concepts. By comparing the static data,the paper tells the concepts of data stream,and it's development process and characteristics;And also,it introduce some current model constructing algorithms of data stream, and association rules and frequent partten mining in data stream; This article introduce current developments and characteristics of data stream management system.2, The paper designs a summary model based on CAN - tree structure. Through introducing a thought of trains, use the data stream itemsets constructing basic orderly table, this improves the compression ratio of the CAN-tree; Improving the tree's structure, to make it conforms to the need of lately frequent pattern mining;3, FPMC algorithm proposed. Based on the improved CAN-tree's structure, mentions a fast algorithm for frequent pattern mining.It makes subsequent mining process saving as many resources as possible,and improving the mining speed and efficiency, make it more accord with dynamic data stream mining thoughts.Generally speaking, through the experiments proved, The design of the system meets the original requirements.Realized a completed application system which accord with data stream mining system defined.