Towards a Highly Configurable Data Mining Middleware through Metadata Approach for Access to Unlimited Data Sources

Abstract

Information stored in a consumer database has become a valuable asset for an organization in today’s information age. It houses vital, hidden information that can be extracted using data mining tools to solve real-world problems in engineering, science, and business. However, due to the rapidly increasing data, data mining tools face substantial challenges in extracting information from data sources such as XML, relational databases, flat files, spreadsheets and so forth. Worst still, these data sources might reside on different locations across different boundaries. Therefore, in order to support new data sources, data mining tools need to provide the flexibility to access the wide range of data sources through straightforward configurations. In this paper, we provide an overview of a proposed data mining middleware known as Java-based Data Mining Middleware which extensively makes use of such configurations through metadata and XML to access unlimited data sources that reside on different locations. Additionally, the proposed middleware also supports a wide range of data mining techniques through trouble-free configurations. Additional data sources and data mining techniques can simply be plugged into the middleware programmatically and non-programmatically.