Improvement of attribute-oriented induction method based on attribute correlation with target attribute

Attribute-oriented induction (AOI) is one of the classical knowledge discovery methods for a relational database query in the field of data mining. On the basis of deeply analysis on the principles of the AOI method, this paper points out some problems existing in it such as redundant attributes after generalization and the invalid rules. This paper puts forward the concept of correlation degree with target attribute, and then gives the improved algorithm according to it Removing the redundant attributes with weak correlation degree with target attribute could help the improved AOI overcome the problems existing in the classical AOI method, and thus improve its efficiency. Different approaches to calculate correlation degree with target attribute are defined to deal with different type of data. Grey relation and attribute reduction based on rough set method are induced to fulfill the above calculation. Experiments on an example demonstrate the effectiveness of the proposed method.