Research On Prediction Of Consumption Tendency Based On Data Mining

At present,the rapid development of e-commerce,online shopping has become a mainstream way of spending,compared with offline,online spending broke time,geographical restrictions,so that buyers and sellers do not need to face each other,can easily reach a deal.And it’s the wealth of product information and the greater variety of purchase options that make consumers spend so much time and effort looking for what works for them.At the same time,due to the virtual nature and randomness of online shopping,it faces great risks.Therefore,how to improve customer satisfaction through effective marketing strategy,and then enhance the competitiveness of enterprises has become an important issue.How to locate the corresponding group effectively from many consumers and formulate a more targeted marketing plan is the next step in the competition and development of online malls.And consumers generate huge amounts of behavior data in online shopping malls,which provides the possibility for analysis to explore consumers’ purchasing intentions and habits.This paper uses real data from an online mall platform of a department store to predict consumers’willingness to buy among the products they interact with.This paper firstly combs through the previous research literature on consumer purchasing intention forecast and analyzes the basic patterns of consumer purchasing intention forecast on the online platform.Secondly,by preprocessing the original data and exploring its basic distribution law,it can provide reference and basis for feature extraction and algorithm selection.Samples are then selected,and the problem of large gap between positive and negative samples is dealt with by means of sliding window structure sampling and random sampling without return.Thirdly,we can construct feature groups in many dimensions by feature engineering,and obtain derivation features that match the practice.Finally,the model is set up,different algorithm models are used to forecast,and the model parameters are optimized in the process to obtain the optimal prediction model.Logical regression and GBDT were used to construct predictive models and validate them on test sets.In order to improve the performance of the prediction model,which is characterized by the output of logical regression,the prediction effect was improved after re-prediction into GBDT model,which is related to the strong classifier based on the regression tree.Finally,a strong classifier based on GBDT and regression tree is proposed.The feasibility of this method is verified by analyzing real data and consumer behavior data.The forecast result of this model has practical significance for online mall upgrade service,accurate marketing to facilitate transaction,upgrade transformation,etc.