Abstract: A method and system is used to predict a behavioral value for a forecast period based on historical pattern identification and modeling. A set of training periods is compiled, and for each of the training periods, a generalized percentage function of the behavioral values is defined and computed. The training periods and their corresponding generalized percentage functions are further normalized and generalized, respectively, so that all training periods have the same length as the forecast period. A prediction time series is computed from the generalized percentage functions. The prediction time series is used to forecast a set of values in the forecast period based on a current behavior value. In one example, the behavior values are revenue and calculated over monthly training and forecast periods.