Identify Potential Knowledge Paying Usersfrom The Perspective Of Network Community

With the popularity of mobile Internet and the maturity of online payment technology,online knowledge payment model has gained wide recognition and attention.This model brings more opportunities to improve the liquidity of the knowledge community and promote the business development of the community.Even so,the percentage of paying users is still low,and more effective identification of potential paying knowledge users is critical to the sustainability of online knowledge communities.At present,most of the researches that examine the influence of knowledge-paying behavior are based on the influencing factors of the product itself,such as price and quality,and ignore the influence of the network community on users’ knowledge-paying behavior.In addition,these researches are based on the user data of small magnitude and lack of universal verification.Some studies have proved that users’ social behavior on social network platforms is an important predictor of users’ paying behavior,but there is a lack of research on how online social behavior affects users’ paying behavior from a deeper perspective.Therefore,from the perspective of network community,this paper explores how community trust,community interaction and community recognition affect users’ knowledge-paying behavior,and the role of these dimensions in the identification of knowledge-paying users.Based on consumer decision theory and network community theory,this paper constructs a theoretical model to identify paying users of social q&a community knowledge from three dimensions: community trust,community interaction and community recognition.Moreover,Zhihu.com(www.zhihu.com),the largest online knowledge community in China,was used as the specific background for the research,and 4 million user behavior data were obtained as the basis.First of all,this article is based on random forest algorithm to calculate the different types and different levels of interaction influence on user knowledge payment behavior,and then,based on logistic regression algorithm characteristics of each variable to explain the impact of knowledge payment behavior,finally,in order to balance the random forest algorithm and high performance and interpretability of logistic regression algorithm,using logistic regression algorithm based on the characteristics of the data set is forecasted.In the process of model training and analysis,this paper optimizes the model by using cross checking method and grid search algorithm,and visually evaluates the three models by various indexes.The results show that community interaction and community acceptance play a more significant role in the prediction of knowledge paying users.The size of feature variables and the importance of feature in prediction are not simply monotonically increasing or monotonically decreasing.In predicting different categories,the importance of features also varies.In general,the active social interaction of users plays the most important role in the prediction of knowledge payment behavior.In addition,the performance of logistic regression algorithm based on feature selection is very close to that of the classifier of random forest algorithm,and the AUC value reaches 0.77.