This study suggests an algorithm for splitting the values of dependent variable (class) while maximizing the information gain about numerical and nominal attributes. Rule induction has been proposed as a way to speed up an acquisition of knowledge for development of expert systems, especially it is a method of automatically developing rules from sets of examples. For market segmentation, the original ID3 algorithm must be modified since it does not exist intermediate stopping rule. Thus, this study suggests an adequate stopping rule. The modified algorithm contains the pruning measure based on Shannon``s entropy and the measure of total information gain. This algorithm is applied to market segmentation problem in medium-large level computer market, in order to develop an adequate market strategy. The derived rule-tree, that transformed in the form of a production rule, classifies the competitive products and in the several market segment well.