Attribute reduction based on the grey relational analysis and dynamic programming

Attribute reduction is an important method in improving the efficiency and the quality of data mining, especially for high dimension data. Therefore, the reduction of data dimensions shall be achieved with higher efficiency. There are many multi-feature attributes in the real project. Many existing works are devoted into finding optimal reducts. In these kinds of methods, the feature attribute importance is always needed in order to do attribute reduction. A grey system has a level of information between black and white. Grey relational analysis (GRA) is a multi-attribute decision making (MADM) method that can be used to find the relationships between one major sequence and the other sequences. In this paper, a method combined grey relational analysis (GRA) with dynamic programming is presented. Considering both the relationship between individual feature attributes with the target attribute and the relationship between different feature attribute, the relational data in our paper include two parts: one is the grey relational grade between ith and jth attribute, the other is jth attribute's contribution to the aim data. Dynamic programming is proposed to reduce attributes of low interest, which can be used to find the best feature in the residue feature attributes with the relational data. Based on the longest length in the dynamic programming, we can easily get the importance of attribute sorting and do attribute reduction. The measure of the attribute importance will be more rational. The proposed method will play an important role in the attribute reduction, data mining, decision analysis and so on. Quality data will be supply for the further processing subsequently.