A New Relationship Form in Data Mining

Abstract

In this paper we study the problems pertaining to the rules mined from the data that do not always hold in the real world. We argue that the cause-effect relationships are more complicated in the real world than those can be presented by the rules mined using the current data mining techniques. Inspired by the concept of cause-effect relationships, the characteristic of the catalyst in Chemistry, and the theory of Net Force in Physics, we propose a new form of representation of rules by introducing a complex cause-effect relationships, the importance of the factors, and force unit. This form of relationship among attributes consists of all of the related attributes and the number of force units of each attribute and also the degree of importance that is like weight of each attribute to the target attribute. The target attribute of interest results from both the change in direction and the number of force units of the change. We have to consider the net force calculated from all of the related attributes including their degree of importance and the number of force unit on the target attribute.