Abstract: A new method to forecast customers' value is put forward using such data mining techniques as clustering and classification. The indicators reflecting old customers' value and business integrity are gained through analyzing the historical transaction data. Then, these old customers are clustered and further classified into different groups in accordance to their value indicators, i.e., each and every old customer is assigned with a mark equivalent to its value. The naive Bayesian classification method is used to forecast new or potential customers' value, and a relevant customer development strategy is thus available. A numerical example is given to verify the effectiveness and practicability of the method proposed.