Accurate forecast is helpful for the enterprises to make decisions，including the production plan， the pricing and promotion decisions，etc． ，so as to reduce the inventory cost and improve the service qualities． In this paper，by analyzing the factors influencing the demand for fast moving consumer goods (FMCG) ， a synthesized forecasting model，which involves both the time series and the multi-regression methods，is established． The synthesized forecasting model is further integrated with the inventory decision，with the purpose of minimizing the overall logistics costs． To solve the integrated model where multiple parameters are involved， a variable neighborhood search (VNS) based algorithm is developed． To evaluate both the synthesized forecasting model and the integrated model of demand forecasting and inventory decision，computational studies are conducted based on some real data． The computational results show an outperformance of the synthesized forecasting model regarding forecasting accuracy，and an outperformance of the integrated model of demand forecasting and inventory decision when the logistics costs are minimized．