The Research Of A Distributed Data Mining System Based On Web Services

Data mining is becoming a key technique in discovering meaningful patterns and rules from large amounts of data (Berry and Linoff, 1997). It is often used in the fields of business such as marketing and customer support operations. Application of data mining is little while research of data mining is very hot. In China, few companies use data mining to make commercial decision. There are some problems in application of data mining: l.Datamining (DM) has another name: Knowledge Discovery from Database (KDD). In China, most databases are small databases, knowledge discovery from those databases is uneconomical and knowledge discovered from separate small database is unbelievable.2. Enterprise Application Integration (EAI) is fashionable now, that can easily resolve the former problem: different companies share the information for data mining. Relative to separate data mining tool, data mining integration system is economical and believable. Now, the problem is how to integrate the application for data mining, while different companies' information systems have different interface, framework and software edition.3. The large company who uses distributed large database has no problem as before, but a new problem is brought: commercial databases are vitals of the company, they are protected by firewall. Now, how distributed data mining system easily get an access to these databases.This paper gives a new data mining model to resolve these problems: distributed data mining system based on Web services.Firstly, the paper gives a data mining model based on Web services, which is the down to date technology. Secondly, a distributed data mining system based on Web services is given. In this part, the paper focus on an implement of Apriori algorithm on distributed circumstance. Then, aim at Internet, this paper gives an optimization for distributed data mining algorithm. Thirdly, after discussing flows of usual preprocessor for Web log mining, a new integral and efficient preprocessor is givenAt last, this paper gives a prototype system based on the forenamed model to mine association rules in Web usage data, which proves that the distributed data mining model based on Web services is efficient, believable, secure and feasible.