Research And Application Of Multi-attribute Classification Based On Case Based Reasoning

As a new research direction of multi-attribute decision making theory, the research of multi-attribute classification has been paid much attention by many scholars in the world. From the point of view of classification algorithm, the classification method of Case-based reasoning(CBR) requires only the decision makers to provide a typical case of decision making and the decision maker's preference and the classified parameter information can be obtained by case study. While direct classification method requires decision makers to provide specific classification parameter information. Decision making based on case reasoning is easy to understand and close to human's daily decision making process which has a wide range of applicability. And the key to the research of this kind of method is how to infer the classification parameters from the existing case classification information. From the point of view of classification results, multi-attribute classification is classified into two categories, which are preference classification and nominal classification. In this paper, we use rough sets and case-based distance methods to extend the multi-attribute classification research based on CBR.The main research contents are as follows:(1) For preference classification problems, a typical case selection strategy based on K-means method is developed and a CBR classification method based on rough set theory is proposed. Then, the two dimensional and the three dimensional classification models are used to solve the two problems of enterprise performance classification and the classification of graduate students' learning ability.(2) For nominal classification problems, a nominal classification model based on CBR is proposed. By calculating the distance between the typical case sets and the group center point, a optimization model of the attribute weights and the classification threshold is constructed, and then the classification parameters are obtained by case study. Which Simplifies the complex comparison process of various kinds of preference information in decision making. The existing model is used to solve the problem of preference classification, this paper extends the model to solve the problem of nominal classification, and applies to the specific case of industrial production group classification.(3) A classification decision support system based on the classification model constructed in this paper. And a case of the fourth chapter is used as a calculation example.