Sales Forecast And Sales Management System Development For Lentinula Edodes Logs ——Take Qihe Biology As An Example

In view of the problems of Shandong Qihe Biotechnology Co.,Ltd.（hereinafter referred to as: Qihe Bio）,the sales forecast of shiitake mushroom sticks depends on experience,the prediction accuracy is poor,and the level of informationization of the sales management of shiitake mushroom sticks is low.,Shandong Province’s major scientific and technological innovation project-R＆D and industrialization of key technologies for smart factory production of edible fungi,using multiple linear regression,machine learning models and other methods to build a sales forecast model,and developed sales forecast and sales management of mushroom sticks The system provides scientific guidance for Qihe Bio in the mode of sales and production,which fundamentally avoids the problem of insufficient supply of shiitake mushroom sticks or overstocked inventory,and reduces losses.The main research contents and results are as follows:（1）Collection and processing of sales data of shiitake mushroom sticks.First,process the15,330 sales order data of Qihe Bio in 2021,filter the varieties with too few orders,and keep the varieties with more than 100 orders;then use python crawlers to obtain data features such as weather,holidays,weekends,etc.The sales data are fused to generate the experimental data set;then the multiple linear regression analysis is performed on the sales influencing factors in the experimental data set,the strong collinearity between the influencing factors is excluded,and the sales factors affecting the sales of mushroom sticks are determined;Factors are processed by feature engineering,including outlier removal,missing value removal,and category feature conversion.After one-hot encoding conversion,the dimension of the experimental data set is expanded from 6 columns to 94 columns.Through the collection and processing of the sales data of shiitake mushroom sticks,the data basis for establishing the model is provided.（2）A forecasting model for the sales of shiitake mushroom sticks based on the combination model was constructed.By analyzing the sales data set of mushroom sticks,combined with the characteristics of the sales data,a prediction model for the sales of mushroom sticks based on ensemble learning was proposed.First,three single models,XGBoost,Light GBM,and Gradient Boost were selected for research.After model tuning and model testing,the average absolute percentage error of Gradient Boost was 2.77%,the average absolute percentage error of Light GBM was 3.71%,and the average absolute percentage error of XGBoost was 3.32%,Gradient Boost has the smallest error in a single model,and the model has the best performance;based on model parameter tuning and model testing,three single models are integrated through the Stacking method to learn the Stacking combination model.The experimental data show that the average absolute percentage error value is 2.07%.Compared with other models,the error value is smaller and can more accurately predict the sales of shiitake mushroom sticks.（3）Design and implement a sales forecast and sales management system for mushroom sticks.Based on Spring Boot,My Batis,My Sql and other technologies,the sales forecast and sales management system for mushroom sticks was designed and implemented,and the data entry,maintenance,retrieval,statistical analysis and sales forecast of mushroom sticks were realized.By realizing the information management of the sales process of shiitake mushroom sticks,the connection between the production data and sales data of shiitake mushroom sticks is established,which ensures the accuracy and availability of sales data,provides a solid data foundation for sales forecast,and provides a solid foundation for the company’s shiitake mushroom production.The sales of mushroom sticks provide a scientific theoretical basis,reduce the situation of insufficient supply or overstock of enterprises,and improve the competitiveness of enterprises.