[Objective]Benign group interaction has a positive guiding role in the process of spreading of fake news. To give full play to the inhibitory effect of social media user's group interaction quality on the negative impact of fake news, and accurately determine the causes and ways of benign interaction, an interpretable model of fake news' group interaction quality based on RF-GA-XGBoost and SHAP is proposed.

[Methods]Taking 500 fake news and 7029 comments from the dataset Weibo21 as the research object. Firstly, the fake news' group interaction quality is comprehensively measured from three dimensions:content, form and emotion of comments. Secondly, the fake news text features are extracted from these three dimensions. Then, the sequential forward search strategy of random forest is used to extract the optimal feature subset of fake news text, and a group interaction quality prediction model based on GA-XGBoost is constructed, and conduct experimental comparisons with other mainstream machine learning algorithms such as LR, SVM and XGBoost. Finally, the SHAP model is used to provide causal explanations for the impact of important features on the group interaction quality.

[Results]The experimental results show that the F1-score and AUC values of the GA-XGBoost model are both above 86%, and the selected six performance indicators are all superior to their comparative models. In addition, the characteristics of false news texts, such as the number of content characters, the number of words, the number of negative emotional words are important factors that affect the fake news' group interaction quality among social media.

[Limitations]This paper does not conduct multi feature interactive interpretation analysis, nor does it dig into the early high-quality group interaction rules according to the timestamp.

[Conclusions]In summary, this interpretable predictive model can accurately obtain the impact of each feature on the group interaction quality, which is conducive to providing effective decision-making support for improving the operational strategy and functional design of social media platforms.