Research Of Big Data Analysis Of E-commerce User Behavior Based On Spark

With the coming of the 21 st century,the rapid development of e-commerce Internet has been involved in every corner,thus causing the growth of e-commerce user data.In order to maximize the profit of the store,the merchants have made many measures,including the analysis and research on the behavior of e-commerce users,so as to judge the purchase behavior according to the behavior characteristics of e-commerce users.In the daily life of Internet shopping,all kinds of casual behaviors of netizens will lead to the final purchase behavior,but few of these behaviors will turn into purchase behavior.Building a set of machine learning model based on users' personal information and online shopping behavior can promote the purchase behavior of e-commerce users.On the one hand,it can improve the purchase efficiency of e-commerce users;On the other hand,it can increase the overall income level of sellers.In order to improve the purchase rate of e-commerce users,the key lies in the e-commerce platform and the recommendation algorithm model.Therefore,this paper mainly conducts the following research:1?Firstly,the research background and significance of e-commerce user behavior are introduced.On this basis,relevant technologies of e-commerce user behavior analysis are studied,including components of Spark platform and relevant classification algorithms of user behavior analysis model.2?Secondly,A series of data preprocessing is carried out for user behavior data provided by an e-commerce platform for a period of time,including the removal of outliers,the processing of missing values and other preliminary work,and then the rules of time series are proposed for the dynamic sliding window processing of the original data.3?Finally,The function and significance of the three parameters of the XGBoost model were analyzed,then the construction of a single model of user behavior and the design of the improved mixed model were described,and the Spark-XGBoost was usedtogether to optimize the important parameters.Finally,the advantages and feasibility of the improved model were verified through comparative experiments between the improved Spark-XGBoost model and the traditional machine learning method.Through the above research,XGBoost prediction model which is based on Spark platform has better prediction accuracy and stability than the traditional machine learning algorithm model.The parallel method for business user behavior prediction is provided by this study.What is more,it can be applied to daily life as an effective way to predict e-commerce behavior.