Font Size: a A A

The Research And Relization Of Mining Frequent Patterns On Business Data Straems

Posted on:2009-05-13Degree:MasterType:Thesis
Country:ChinaCandidate:R GuFull Text:PDF
GTID:2178360245986018Subject:Management Science and Engineering
Abstract/Summary:PDF Full Text Request
With the advent of the knowledge economy era, information and knowledge has become an important strategic resource and the core competitiveness to an organization and a nation, and also the foundation in the implementation of scientific management and decision-making. Therefore, how to gain information and discover knowledge especially in the dynamic and explosive growing data streams become the key issues.Different from the traditional data, the data stream is abounded, rapid, and continuous. In addition, the business data stream is continuous, conflict, timing,massive and distributed, so traditional data mining techniques can not be applied directly to the business data stream. Making use of the limited system resources to obtain useful information from the business data streams has brought new opportunities and challenges for the application research of data mining in business areas.Frequent patern mining is a basic problem of data mining, including mining transactions, sequences, trees and graphs. The algorithm for it has been prevalently used in many other data mining task, such as association analysis, sequence-period's analysis, maximal and closed frequent paterns, query and classification technology etc. Since it lays groundwork for other problem and its intrinsic complexity, the algorithm for frequent patern miming has become the focus of many research workers. Some relevant techniques about frequent pattern mining in the business data stream are addressed in the thesis, which covers the analysis of the level and structural dimensions of the business data stream, mining maximal and closed frequent patterns by using the static gleaned tree efficiently, using incremental mining methods and the tilt time window respectively to mine the maximal and closed patterns in the business data stream, the application of frequent pattern algorithms in the business field. Major contributions of this thesis include:Firstly, study the relevant theories of the data stream mining and its model, summarize the latest research achievements of the field in order to use them in the business data stream mining tasks.Secondly, extract concepts and features of the business data stream, analyse the content hierarchical structure and type dimension structure and use them to construct the Business Data Stream Management System BDSMS.Thirdly, according to the characteristics of static business data, design the Maximal Frequent Pattern mining algorithm MFP and the Closed Frequent Pattern mining algorithm CFP. Use the feedforward pruning, the merger strategy to enhance the construction of the frequent pattern. On this basis and response to the time series model and the cash register model, bring in the incremental mining and the tilt time window respectively to get the single scanning algorithms SMFP and SCFP.Finally, use the above algorithms in the specific business areas, and design a retailor discount coupons generation system. The experiment results show that the algorithms have high accuracy and time efficiency, which can be used in the business decision making support.
Keywords/Search Tags:data mining, datastream, frequent pattern, maximal frequent pattern, closed frequent pattern, incremental mining, tilted-time window
PDF Full Text Request
Related items