Order Forecast Of Tobacco Distribution Center And System Implementation

Tobacco Industry is meeting the openness of the markets and is promoting the reform of "Supply by Orders". In this background, it becomes most important for Tobacco Company to forecast amount of orders and master the demand of the markets. Forecasting orders of Tobacco Distribution Center can provide data support to ascertain ordering point and ordering quality, furthermore, reduce the cost of stock, improve the fulfillment rate of customers' orders. Making Tobacco Supply Chain effectively organize purchasing, producing and distributing, and improving total effect.The thesis forecasts custmers' orders of Tobacco Distribution Center by period, brand, district, from several dimensions. When forecasting year's order amount, using Regression Models, Grey Model, Self-adaptive Secondary Exponent Smooth Model and Moving Average Model, seven models. The mean accuracy of Logarithm Regression Model is highest, gets 98.45%. Grey Model GM(1,1),Regression Based-on Gross Local Product Model, the accuracies of forecasting results using Self-adaptive Secondary Exponent Smooth Model, Linear Regression Model, Exponent Regression Model and Moving Average Model are 98.25%, 97.16%, 96.85%, 96.21%, 95.6%, 84.97% respectively.When forecasting month's order amount of tobacco, the thesis uses four-level BP (Back Propagation) Neural Network (NN) to forecast, meanwhile, uses Genetic Algorithm (GA) to optimize the weights of neural networks, overcomes the shortcomings that Neural Networks apt to be trapped in local optimum when searching values of weights. This algorithm speeds up converging, improves the ability of generation of networks. The Neural Network gets the expected accuracy (98%) after training and testing. When forecasting month's order amount, the thesis also uses Regression Models, Grey Model and Self-adaptive Secondary Exponent Smooth Model, The mean accuracy of Linear Regression Model whose effect of forecasting is better is 96.9%, Self-adaptive Secondary Exponent Smooth Model applicable to forecast total tobacco, tobacco inner province and tobacco outer province, gets the accuracy which is 96.81%, 97.12%, 85.6% respectively.Considering the practical work period of Tobacco Distribution Center and the seasonal influences, when forecasting day's order amount, the thesis introduces PROPERTY_TAG to revise the forecast, improve the accuracy of forecast and easy to realize programming. Based on this, the thesis uses Proportion Model Based-on Lunar Calendar, Regional Trend Model, ARMA Model and GA Optimized BP Neural Network Model to forecast. GA Optimized BP Neural Network Model gets the expected accuracy which is 98%, but it will take much time as training samples and evolution generations increase. Proportion Model Based-on Lunar Calendar and Regional Trend Model can reflect the practical conditions of Tobacco Distribution Center, their accuracy are 90.87%, 95.61% respectively. The mean accuracy of ARMA Model is 96.61%.When forecasting week's order amount of tobacco, using Accumulation of Mean Order Amount per Day Based-on PROPERTY_TAG and Regional Trend Accumulation Model.The former whose accuracy is 94.15% and cost time is 14s, is superior to the latter.This thesis develops Order Forecast Web System of Tobacco Distribution Center based on J2EE platform, applying Struts framework which can separate view and logic, easy to improve models, and strengthen the reusability of the program. The system has friendly human-computer interaction, its forecast results are correct and intuitionistic, and provides date support to decision makers.Meanwhile; the system has extensibility, and can further improve Inventory Control System, build the coordination information platform of Cigarette Industry, and integrate all resources of Cigarette Supply Chain.