Study On Prediction Of E-commerce Consumer's Buying Behavior Based On Data Mining Technology

Customers are the source of life for an enterprise,and customer resources with a high conversion rate are the guarantee for the steady development of an enterprise.The rapid development of Internet information-related technologies and the overall popularization of smart phones have made Chinese e-commerce business model more and more mature,and the era of e-commerce has arrived.The increasing interaction between buyer and seller has caused e-commerce enterprises to face various competitive pressures while enjoying technological dividends.The core issue that enterprises have been discussing is how to more accurately predict customers' needs and improve customers' conversion rates.Under this circumstance,using existing resources to improve the accuracy of customer demand forecasting is an important factor for e-commerce enterprises to enhance their competitiveness.In recent years,as the development of science and technology has become a global consensus,the maturity and application of many high-tech technologies has led to an exponentially rapid growth in data volume.At the same time,the advancement of data mining technology enables big data to truly become a resource that enterprises can effectively use.Big data is a powerful tool for e-commerce enterprises to acquire customer needs and achieve precision marketing.Compared with traditional precision marketing methods such as telephone and email,precision marketing through big data and data mining technology has the advantages of high efficiency and low cost,and it reduces blindness and subjectivity in the marketing process.Typical application fields are the Internet and e-commerce and other emerging fields.To achieve enterprises precise marketing goals,this study uses consumers' historical data to model and predict consumer demand through data mining technology.The research content is to use data mining methods to process consumers' historical data of an e-commerce enterprise.In the step of feature engineering,construct labels and feature variables according to the business needs,and then use representative regression algorithms to model and analyze the processed data set.At last,evaluate and compare the performance of different regression prediction algorithms.The main conclusions are:(1)Building labels independently according to the needs of the problem when the data set doesn't give labels,and the constructed labels should be reasonable and meaningful.In the process of feature construction,it is necessary to make full use of the original data.Determine the direction of feature construction by reading related professional field literature,and select features to make the model more refined.(2)In feature selection,two different methods are used,one of which is a combination of filtering and embedding,and compared with the feature recursive elimination algorithm in the wrapper method.The results show that for the data in this article,the feature set extracted by the first method is more effective.(3)Using linear regression,Cart regression,KNN,Lasso regression and Light GBM to model the historical data of e-commerce users.Through the performance evaluation indexes such as RMSE and S1S2,it is found that these regression models can effectively improve the prediction effect and the linear regression model has the best performance.