To dynamically extract a suitable category and to grasp the most approximate demand trend by selecting a category capable of recognizing the demand trend and deciding demand forecasting quantity based on the demand result of the selected category.

A forecasting object commodity name whose demand is to be forecast is obtained by an input device 114, and the factor and weighting of the forecasting object commodity are obtained from a commodity factor DB 100. The taken out factor whose content by individual commodity is equal to that of the forecasting object commodity is selected and is extracted as one category in a category classification DB 102. The real demand for an arbitrary period of the commodity belonging to the extracted category is obtained from a real demand DB 104 by individual products, they are time-sequentially summed up and the real demand of the category is calculated. Multiple regression analysis is executed on the calculated real demand and it is judged whether the demand trend is recognized or not. When the demand trend can be recognized, the demand forecasting quantity at the demand forecasting object date of the category is calculated by regression analysis and demand forecasting quantity is decided.