Design And Realization Of The Model Of Semiconductor Components’ Sales And Purchases Related Forecast

To improve prediction accuracy, reduce prediction error range is a common goal for most prediction theories and methods. When multiple predictions relate to one another, uncertainty diffusion may become very complicated, in some occasions a variety of forecasting uncertainty may be reduced by offsets, in other cases more uncertainty may be aggregated, which makes the prediction uncertainty further spread due to their inner relationship. The thesis was focused on the related predictions of sales and purchase, considering the transferring diffusion effect of the uncertainty between the two different related predictions. The model of related predictions of sales and purchase was conducted using interval values. And the results were used as the input of a new inventory control model.The related prediction model proposed in this thesis firstly used the gray volume and liner programming for sales forecasting result with its intervals. The result played as an input of MRP operation, which took into account of the data error from BOM and a variety of constraints from inventory, capital and production. Using mid-point and width of interval as controls, the model optimized the objective function in two different ways both of which utilized the particle swarm algorithm. Finally an inventory control model was built to verify the availability and feasibility of the prediction model.Compared to the traditional prediction model of sales and purchases which was based on single-value inputs, the related prediction model in this thesis could quantitatively describe and optimize the transformation process between sales forecast and purchase prediction through the form of interval values, the model could also effectively control the uncertain diffusion problem by algorithms of interval number. On the one hand, the results of the inventory control model verified the correction of the related predictions; on the other hand it served as a supplement and extension of the related prediction model.