We are glad to share with you that our recent international journal paper “Optimization of Futures Price Forecasting and Trading Strategies Based on Clustering and Multi-model Fusion” from our MSAAI student Zhicong Song, which has been published in MDPI Journal of Forecasting (JCR Q1):
- Zhicong Song, Harris Sik-Ho Tsang, Richard Tai-Chiu Hsung, Yulin Zhu, and Wai-Lun Lo, “From Market Volatility to Predictive Insight: An Adaptive Transformer–RL Framework for Sentiment-Driven Financial Time-Series Forecasting,” Forecasting 7, no. 4: 55, October 2025. https://doi.org/10.3390/forecast7040055
Paper Abstract
This study addresses the challenge of predicting financial market trends, which is difficult because markets change quickly and are influenced by both data and how investors feel. Traditional methods often miss important information from investor opinions shared online, while many AI approaches can’t adjust well to changing market conditions like bull or bear markets.
We developed a new system that combines investor sentiment from online forums with an adaptive AI technique that learns and adjusts over time. We built a special financial sentiment dictionary with over 16,000 words to understand forum discussions accurately, reaching 97.35% accuracy in classifying forum topics. We then used both historical price data and investor sentiment with advanced AI models to predict market movements. Another AI agent actively combined these predictions to make smarter decisions.
Our tests on various markets—including stocks, commodities, and indexes in China and US—showed that this hybrid approach works better than existing methods. It’s more reliable during market ups and downs and performs well across different types of assets, providing a promising way to improve financial forecasting using AI and investor sentiment.
Some photos of the authors and papers
Overview of our proposed hybrid framework architecture.
Model prediction across time on CSI 100 index.
