Abstract: Identification of customers with potential market value is a key problem in direct marketing. Although some standard data mining methods may be applied for the purpose of direct marketing, no ideal results have been achieved, Therefore, based on the informational theory, the market value algorithm，is adopted to propose a linear market value function so as to discover the customers with the most potential market value. The experiment result shows that its evaluation value can be up to more than 80% and this algorithm is well interpretable. The authors try to improve the predictive ability of market value algorithm by applying Boosting to it. Experiments are carried out on three different data sets and the result shows that one percent point of evaluation value can be improved on average.