Research On Customer Lifetime Value Based On User Reading Behavior Characteristics Of News Industry

In recent years,with the rapid development of the Internet economy,the competitive environment of enterprises has become more and more fierce.The competitive strategy of enterprises has gradually changed from product or service in the past to customer-centric.So how to acquire and maintain high-quality customers has become the key to the survival and development of enterprises.Therefore,how to correctly identify high-quality customers has become an important direction in studying enterprise customers.At present,the research on the lifetime value of users is mainly concentrated in traditional industries,such as telecommunications,retail,banking,insurance,etc.,but there are few studies on the lifetime value of Internet users,so this paper takes the Internet news industry as an example to study it.The customer lifetime value is the value contributed by the customer to the company's business throughout the life cycle.This paper combs the theoretical model of the customer's lifetime value and finds that the basis of the current research is the net present value model,so this paper also draws on the model to give a little information.Observing and pre-processing based on real user data in 2017 and 2018,and then using factor analysis to reduce the dimensionality of multiple user behavior indicators to determine key features in the reading characteristics of news and information industry users,combined with the news industry Realizing the commercial benefits by means of advertising monetization,finding the relevance of advertising revenue and different reading behavior characteristics of users,and then expressing it as the benefit brought by the users in the net present value model,establishing the news information industry based on user reading A measure of the value of behavior.Then,this paper studies the related theories of machine learning algorithms,and uses the GBDT and XGBoost models in the machine learning algorithm to classify the behavior data and value data of the users in 2017 into training sets and test sets for model training.Through step-bystep iteration to find the optimal model parameters,which can be used to predict the value of current users in the coming year,and guide enterprises to make marketoriented decision-making and customer acquisition costs for user growth.Through the research in this paper,it is determined that the user's information flow browsing depth in the user's reading behavior characteristics,that is,the number of information flow refreshes,the reading depth,that is,the total number of article readings,and the user's willingness to revisit,that is,the total number of user open applications constitute the key behavior characteristics of the user.The commercialization coefficient related to the three behaviors of the user finally represents the benefit that the user brings to the enterprise in the net present value model of the user value,and establishes a model.Then compare the prediction accuracy of GBTD and XGBoost two machine algorithm models,and finally choose XGBoost model as the user lifetime value prediction model.In practical application,in general,enterprises will take the predicted value as the benchmark,further adjust the fluctuation range of its value,calculate the range of cost-benefit ratio,and then determine the quality of the channel to decide whether to use the channel to launch.