Graph Transformer for Multi-Stock Return Forecasting

This presentation explores predicting 7-day stock returns in interconnected financial markets using cutting-edge Graph Transformer models. It covers data preparation from sources like S&P 500 stocks and sentiment analysis, addressing challenges like incomplete data. The methodology focuses on feature engineering, standardisation, and target forecasting. A baseline model is compared to a novel multi-edge graph-based approach. Results highlight improved classification accuracy, with future...