Abstract: Multivariate regression analysis is employed to predict the prices of mechanical and electric products according to historical prices. Because there are many years in which the product price data are unavailable, a price index is introduced into the prediction. Thus, the problem of forecasting the product prices is converted into the one of forecasting the price indexes. In the regression analysis, the time and price policies are taken as significant factors of the price index. The characteristics of the fitting values of the price index are also discussed.