Fusing Deep Learning and Fuzzy Logic for Social Media Influence Assessment

Focusing on the field of social media influence prediction, this study aims to evaluate the efficacy of a model that incorporates Long Short-Term Memory Networks (LSTMs) with fuzzy logic. Through experimental comparisons, we found that LSTM combined with fuzzy logic performs well in predicting social media interaction metrics, especially when dealing with complex and uncertain social media data. The experimental design covers datasets from two typical social media platforms, and multiple models are used for comparison, including traditional ARIMA, rule-based models, and deep learning models such as GRU and RNN. The results show that the fusion model not only improves the prediction accuracy but also demonstrates stable performance improvement under different event categories. In addition, we analyzed the association between user behavior and influence, proposed a multi-dimensional approach to comprehensively assess user influence, and tailored influence enhancement strategies for users.