Analyzing and Predicting Online Public Opinion on Films Based on Text Mining and Deep Learning

This paper analyzes and predicts online public emotion during the launch of the Chinese animated film Nezha: The Demon Child's Havoc in the Sea in 2025, for example. Through word frequency analysis and LDA topic modeling, it is found that the focus of public attention evolves over time, focusing on characters and box office during the premiere period, and then shifting to director's IP and commercial performance in the later period. Emotional analysis shows that audience emotion is driven by factors such as box office nodes and holidays, showing a fluctuating pattern of “short-term surge - rapid decline”. Compared with the ARIMA model, LSTM performs better in capturing nonlinear emotional changes and is suitable for predicting short-term public opinion trends. The study helps to understand the evolution of movie public opinion and provides reference for market dynamics and emotional management.