Multi-Source Data-Driven LSTM Framework for Enhanced Stock Price Prediction and Volatility Analysis

Abstract

This study used the long short-term memory network (LSTM) model to predict the multi-source data of Tesla stock. The data features include opening price, closing price, highest price, lowest price, trading volume, and financial indicators such as the Sharpe ratio. By comparing the performance of models such as ARIMA, GARCH, GRU, XGBoost, and Prophet, the experimental results show that LSTM performs best in prediction accuracy, especially in capturing the long-term trend of stock prices and overall volatility. Although the model has a certain short-term prediction bias on the test set, it can accurately reflect the trend of Tesla's stock price overall. The research results show that the introduction of multi-source data and financial indicators can effectively improve the prediction performance of the model and provide new ideas for financial time series prediction. In the future, the structure of the LSTM model can be further optimized, and more financial data features can be introduced to improve the model's sensitivity to short-term market fluctuations and prediction accuracy.