Research And Application Of Cigarette Order Forecasting Based On Improved Random Forests

The "new retail" of tobacco comes into being with the pace of the times.Forecasting cigarette ordering volume helps manufacturers to carry out scientific operation activities such as raw material storage and production,which can help retailers to scientifically arrange the number of cigarettes of various categories and prices in the store,improve the utilization rate of funds,and improve the profitability of the store.In order to further improve the prediction accuracy of cigarette order quantity prediction model based on random forest,K-means is used to cluster the data sets,and then bootstrap sampling and integration are carried out for each set to obtain the basic learner training samples of data balance.Then,the final result of the random forest is to take the mean value of the output of all the basic learners,and the weighted average value method is used to optimize.According to the MAPE of the data outside the bag of the basic learners,the weight formula is derived,and the weighted random forest prediction model is constructed.The results show that the prediction accuracy of the weighted random forest model is improved obviously and the comprehensive performance is better.In order to further improve the prediction accuracy of cigarette order quantity prediction model based on random forest,K-means algorithm is used to process and analyze the training data,and then the similar data sets are randomly sampled and integrated to obtain the basic learner training samples of data balance.Then,the final result of the random forest is to take the mean value of the output of all the basic learners.The weighted mean value method is used to optimize,and the corresponding weight is calculated according to the prediction average absolute error percentage of the data outside the basic learners' bag.The results show that the prediction accuracy of the weighted random forest model is improved obviously and the comprehensive performance is better.Design and implementation of Jiangxi tobacco data statistics system.The system adopts the micro service architecture mode,realizes the monitoring of daily sales data of retail customers,makes a statistical display of various scenarios of relevant data of retail customers and consumers,forecasts the next cigarette order quantity of retail customers,and carries out authority management for users.The system can help manufacturers to understand the cigarette market,help retailers to understand the operation of the shop and order cigarettes with one key,which can be widely used in production and operation.