Method and system for predicting revenue based on historical pattern indentification and modeling

A computer-implemented method for determining a total forecasted behavior value for a forecast period having a plurality of discrete time points, the method comprising:

• reading into a computer memory, a plurality of behavior values from each of a plurality of training periods preceding the forecast period, each behavior value associated with a discrete time point within the respective training period, wherein two or more of the training periods have different lengths;
• generating in the computer memory by a computer processor, a percentage function associated with each training period, the percentage function representing percentages (Rn) of the behavior values relative to a behavior value total within the respective training period, wherein the percentages Rn are calculated from cumulative revenue values of respective multiple first intervals, a total revenue for a second interval including the multiple first intervals, and a value representing a combination of delayed revenue and an adjusted revenue amount for the end of the second interval;
• forming in the computer memory by the computer processor, respective generalized percentage functions by normalizing the percentage functions, each generalized percentage function having a period equal to the forecast period, wherein respective generalized percentage functions having periods equal to the forecast period are formed from the percentage functions associated with the respective training periods where at least two of the training periods have different lengths;
• generating a behavior time series in the computer memory by the computer processor, as a function of the generalized percentage functions at each discrete time point of the forecast period;
• generating in the computer memory by the computer processor, the total forecasted behavior value as a function of a current behavior value divided by a value of the behavior time series associated with a current discrete time point of the forecast period.