Towards the Integration of Constrained Mining with Star Schemas

A growing challenge in data mining is the ability to deal with complex, voluminous and dynamic data. In many real world applications, complex data is organized in multiple inter-related database tables, which makes their analysis as a whole more difficult and challenging. The most used multi-dimensional model in data warehouses represents data through star schemas, that consist of a central fact table, linking a set of dimensional tables, representing respectively the business events and dimensions. There are few techniques dedicated to the analysis of these star schemas, with the aim of finding frequent co-occurrences, or patterns, in data, and both suffer from the lack of focus on user expectations. Indeed, one of the common criticisms pointed out to the pattern discovery task is the fact that it generates a huge number of patterns, independent of user expertise, making it very hard to analyze and use the results. Constrained mining is the most used approach to minimize these problems, by applying user defined constraints to filter and focus the discovery process. In this work we propose the integration of these two important areas of data mining, and discuss how this can be done using the already existing techniques.