Enhancing Stock Price Prediction with Distributed Learning

Machine learning has become a crucial tool in stock price prediction, yet traditional centralized approaches face limitations in scalability and efficiency. This paper explores the application of distributed learning for stock prediction, leveraging Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) models. By conducting experiments under different conditions—varying the number of clients and data distributions (IID and Non-IID)—we evaluate the impact of distributed learning on prediction accuracy and convergence speed. Our results demonstrate that distributed learning enhances training efficiency and maintains high predictive accuracy, though performance varies with data heterogeneity and client participation.