Gradient boosting algorithm integrating video data applied to prediction of mutual fund returns

Roadshow video data from Chinese public mutual funds are used to construct a multimodal feature system incorporating textual semantics, linguistic structure, and vocal behaviors． A gradient boosting regression (GBR) model is used to predict next-day fund returns． Model parameters are optimized through cross-validation and grid search with the 2020 dataset of Chinese mutual funds. Comparative analyses with support vector regression (SVR), random forest (RF), and Lasso regression show that this GBR model achieves significantly higher predictive accuracy． Interpretability analysis further indicates that linguistic and acoustic features proportion of vague expressions, speaking rate, pitch variation make prominent contributions to prediction performance． These findings confirm that language style and communication patterns contain meaningful behavioral signals that affect investor judgments and market responses, offering forward-looking informational value． The present work extends the application of multimodal data in fund analysis, provides quantitative evidence to support fund managers in optimising video-based disclosures and investors in identifying non-financial signals．