Research And Application Of Spandex Product Sales Forecast Technology

Sales volume forecast has always been a key research issue.An accurate sales volume forecast can control the cost of the enterprise,plan the market and deploy the strategy in advance.A good solution to this problem can provide theoretical support for the future planning of enterprises.Spandex is the abbreviation of polyurethane fiber,but it needs to be kept in a constant temperature and humidity environment,with a temperature of 18-20 ?.And should not be placed in the sun,not with sulfur dioxide,nitrogen oxides and other chemical products stored together.These restrictions also make the enterprise in the production of Spandex Products and other products to have a more rigorous production plan,otherwise it will cause a large number of product damage.Therefore,the sales forecast of Spandex Products is of great significance to spandex production enterprises.In this thesis,xgboost,SVR(support vector regression)support vector regression and Gru neural network are used as the basic models,and lightgbm is used as the final prediction model to improve the prediction performance of the model,more close to the real sales data,and provide a new prediction method for regression prediction.In addition,new features are integrated.This paper crawls the data of household stele from the website,and processes it by word segmentation and emotion classification.For historical sales data,a series of features are constructed by using the method of sliding window.Finally,the features selected from the features and the features constructed are fused as the training data of the last level model.In the first mock exam,the first mock exam and the combined model are used to compare the performance.The final experiment shows that the prediction performance of the combined model is higher than that of the single model,and the prediction results are closer to the real sales data.Furthermore,this paper compares the algorithm without new features and the algorithm with new features,and also finds that after the introduction of new features,the prediction accuracy of the algorithm has been improved,which also proves that these constructed features do play a role.In the algorithm application and D enterprise color intelligent sorting system studied in this paper,the intelligent sorting system is briefly introduced,with emphasis on the module of algorithm application.Through the prediction results provided by this algorithm,a reasonable production plan is given,which provides a reliable theoretical basis for user decision-making.