To analyse emotions and predict dissemination trends of Sichuan handicraft intangible cultural heritage (ICH) inheritors, this study proposes an ALBERT-based sentiment analysis model and a TCN-based prediction model.Integrating multi-dimensional feature extraction, attention mechanisms, and a dual-channel TCN structure, the models capture long-term dependencies by fusing emotional features with dissemination indicators.Experiments on social media data show the improved ALBERT model achieves 86.21% precision and 83.97% recall in sentiment analysis, outperforming BERT, RoBERTa, and LSTM while reducing memory usage.The TCN-based prediction model attains MAE of 0.389%, MSE of 0.007%, and 91.36% fit, improving accuracy and stability.Emotional distributions exhibit periodic fluctuations during festivals, with dissemination trends driven by both emotional and behavioural indicators.This research enhances ICH emotion analysis and prediction technologies, revealing group emotion patterns and communication evolution to provide data support for government short-video timing decisions and inheritors' content optimisation.
Lifu Xu (Thu,) studied this question.