GUIDE : B. ROJA SRI ,K. TEJASWI, M. JITHIN VIJAY SAI, L. AKANKSHA, B. PAVAN KUMAR

Abstract

n today’s fast-evolving technological landscape, accurately predicting market trends plays a critical role in minimizing financial risks and maximizing potential returns. This work introduces a novel approach, the MS-SSA-LSTM model, designed to harness multi-source data in the prediction of stock prices. By incorporating sentiment analysis, swarm intelligence techniques, and deep learning, this method analyses data such as posts from the East Money forum to create a custom sentiment lexicon and calculate sentiment indices. These indices are then integrated with traditional market data, with the Sparrow Search Algorithm (SSA) optimizing the parameters of a Long Short-Term Memory (LSTM) network. The results show that the MS-SSA-LSTM model significantly improves forecasting accuracy, achieving an average R² increase of 10.74% compared to standard LSTM methods. Furthermore, the combination of sentiment indices and hyperparameter optimization enhances the model's performance, offeri