Research And Application Of Logistics Stock Control Based On Combined Data Analysis

The market changes quickly and the competition environment is vehement nowadays. In such state the stock control influences not only the cost of the enterprise and the occupy of its funds, but the level of customer service. Therefore it is necessary to take some right measures to resolve the problem of stock control.This thesis is laid in the production and stock of AL Company. It makes an intensive study of its stock control.We examine this problem from three points of view: manufactured goods, semi-manufactured goods and raw materials. We take different measures respectively. The forecast of the sales of the manufactured goods and the analysis of the stock of the semi-manufactured goods are two focuses of this thesis.We adopt various methods of data analysis and forecast to analyze and forecast the sales of its production.Firstly, we adopt the regression method to analyze the relation of the sales of its two kinds of goods.Secondly, we adopt the time sequence method to build a model of ARMA and analyze the historical selling data. We want to find the correlation among the monthly selling data.Thirdly, aiming at the limitation of the traditional stock control method, we adopt the BP Neural Network method to analyze the historical selling data. We carry on some improvements to the traditional BP Neural Network method: a new method of data pretreatment to process the original data standardization; a new method to decide the number of nodes in hidden layer; a dynamic learn-rate. The experiments indicate that the number of nodes decided by the method above is the best value of this thesis. We decide other values of the BP Neural Network.Finally, we build a fussy combined forecast model based on the liner regression method,time sequence method and BP Neural Network method. We use the fussy mathematics method to decide various weights of the methods above. Then we analyze the sensitivity of the model and verify its robustness. The experiments indicate that the forecast model has a preferable robustness. So we can apply it to practice.We develop "the Platform of Combined Data Analysis and Simulation in Stock Control" based on "visual Basic, and we integrate the algorithms mentioned above in this platform. The platform is universal and can be applicable to the stock control of various enterprises to provide effective suggestions.