Backward induction for web mining in data warehouse environment

Abstract

For web mining, the biggest problem is the scarcity of data. All web mining procedures starts with the identification of the needed data items. To overcome the problem and prepare as many needed data as possible for business intelligent information, we propose backward induction procedures in web mining. Backward induction itself emphasizes the importance of data preparation in all stages of web mining. It identifies the needed data items before the mining stages with the given suitable procedures. Web mining itself is an iterative process where data mining techniques are used back and forth and iteratively. To support backward induction and other web mining characteristics, the concept of scalability is very important. To accomodate the scalability needed in web mining process, we propose the reference web mining architecture in data warehouse environment. The referred web mining architecture has three kinds of scalabilities: the scalabilities of operational databases, the scalabilities of data model and the scalabilities of data mining engines. By following the backward induction procedures in web mining process, we can extract the business intelligent information from web mining.