Analysis of the Influencing Factors of Users’ Adoption Behavior in Social Q&A Community Based on Machine Learning Regression Algorithms

Users are the core component of social Q&A communities and influence the future development of the communities. Therefore, it is particularly important to ascertain the influencing factors of users’ adoption behavior. Based on the information adoption model, 41,832 Q&A data are collected from the Zhihu community between May 27, 2011 and March 14, 2018. The data comprises three thematic areas: children education, medical care, and Internet finance. This study explores the impact of 122 features on the adoption behavior. Four machine learning algorithms are selected. They are multiple linear regression, Lasso regression, Ridge regression, and Xgboost regression. By comparing R
²
and RMSE of the models, the optimum model is used to filter the top 20 important features of each theme. The results show that 1) compared with the traditional model, the Xgboost model has better prediction performance, with a maximum accuracy rate of 63.9%; 2) answer content, answer background, respondents, and answer forms all affect adoption behavior; 3) users of different themes pay different attention to the features of each category. These results indicate that clarifying the scope of the themes and distinguishing them can improve models’ prediction accuracy of model. On the other hand, the research also proves the application value of the information adoption model and the Xgboost regression model.