A Novel Approach To Product Precision Marketing In Industry Based On Ensemble Learning

The traditional means of sales in the manufacturing industry is an important means to sale traditional manufactured industrial products,which can be sold for a long time and profitable for enterprises.Traditional manufactured industrial products(such as parts)have some obviouscharacteristics,such as short replacement cycle and large demand for periodic replacement.The traditional sales methods mainly include regular return visits,magazine advertisements,factory bidding information,etc,which not only has high customer cost,low product information transfer rate,but also is not conducive to maintain replacement and sales of parts for customer after-sales service.This paper transforms the precise sales demand target of industrial products into machine learning and data mining problems.Based on the prediction of users' cyclical purchase behavior of industrial products and the prediction of purchase time,this paper focuses on the precision marketing of traditional manufacturing products to study.Finally,this paper designs a user precise marketing strategy method based on integrated learning algorithm and multi-model cross-fusion.The main innovations of this paper are as follows:1)In order to improve the accuracy of the forecast results,this paper abandons the traditional single view of the demand problem and disassembles the precise marketing demand scenario of traditional manufacturing industrial product marketing.Finally we disassemble it into two aspects problems: user purchase behavior prediction and user purchase date prediction;At the same time,based on the multi-type model fusion user model which is based on integrated learningand the model stacking based date model,this paper abandons the traditional single model solution and designs a user precise marketing strategy method based on integrated learning algorithm and multi-model cross-fusion.2)In view of the user's purchase behavior prediction problem,this paper constructs two different types of user models by using the LightGBM algorithm in integrated learning from different perspectives of classification and regression.Then,the user model of the classification model and the regression model are combined by the weighted linear modelstrategy.At the same time,according to the demand scenario,this paper also uses the post-processing on the combined user model.Finallythis paper uses the method to predict the target user group of the purchase behavior;3)In view of the user's purchase date prediction problem,based on the LightGBM algorithm in integrated learning,this paper uses the time-sequential division line verification set and the sliding window periodic multiplication training method to construct the date model.At the same time,based on the correlation between the user purchase behavior prediction problem and the user purchase date prediction problem,this paper uses the integrated learning combined strategy called model stacking to bridge the two problems and cross-integrate the user model and the date model.Finally,this paper realizes the accurate forecasts of user purchase time.The experimental results show that,the proposed method can make a good experimental resultsunder the evaluation indicators which includes AUC(an area under ROC curve)and mean square error(MSE).Finally,this paper provides the profound and significance guidance for the precision marketing of traditional industrial manufacturing products.