A Method on Defining Target Variables and its Application

This paper focuses on a method of defining target variables, which is not defined accurately in data mining and its application. In the area of data mining, there is a class of target variables that can\´t be defined accurately, such as the definitions of high-end customers and degree of customer loyalty in the customer relationship. However, defining target variables is pre-requisite to construct a model of supervised learning. This paper introduces a method, "thermometer-onion", for defining the target variables and presents an example to illustrate its application.