| With the rapid development of the market, the coal enterprise using of the subjective experience is difficult to accurately predict methods its coal sales forecast, to reduce decision-making error, avoid warehouse inventory shortages or the backlog, increase the prediction accuracy of coal sales forecasts, the study of the method is particularly.In this paper a large coal enterprise, the characteristics of the problem, according to the actual with the improved BP algorithm. With the genetic algorithm to optimize its connection weights, and realized the prediction method in coal enterprise sales management application. Thesis mainly completed the following working content:(1) Put forward the neural network method of coal sales forecasts. From coal enterprise actual problems of the sales forecasts start. Through to the coal sales change characteristics and coal sales forecasts the study of the method of analysis. Put forward by using the neural network technology coal sales forecasts method.(2) Established three tiers of the embedded genetic algorithm improved the BP (Back Propagation) network coal sales forecast model. Through to the characteristics of the problem, BP algorithm in weights and optimization process the defects existing in the analysis. Based on the research based on improved BP algorithm, a three layers modified BP network (LM method) forecast model, solve the network slow convergence speed. The detailed design the BP network topology structure, through the test of the comparison method to determine the parameters for the model. By using the genetic algorithm of global searching capability optimize network connection weights, solve the training process easily into the local minimum points. The detailed design the coding method and genetic operators.(3) Put parameters in the model selection in the attempt. Through the analysis of enterprise sales market, combined with the main coal consumption department predicts thoughts of model selection of input parameters in the research on deep.The selection of parameters are covered this enterprise coal of sales in many impacts, and comprehensive reflection of the model accept change information of coal sales. According to the characteristics of the sample data input, data pretreatment methods of analysis and design.(4) Completed the coal sales management subsystem of design and implementation. In the research on the base of coal sales forecasts method applied to coal sales management. The management subsystem of coal sales overall framework, system structure, database, function module design, analysis, and realized the forecast method for enterprise of coal sales management application.Theoretical analysis and experimental results indicate that the neural network for coal of sales of prediction is feasible and effective, has a very good prospects, and the genetic algorithm embedded neural network technology more improve the computational speed and reliability of the prediction. |
