Integrating Deep Learning and Explainable Artificial Intelligence Techniques for Stock Price Predictions: An Empirical Study Based on Time Series Big Data.

This paper proposes an approach to improving the accuracy of predicting stock prices. The approach is built on integrating Long-Short-Term Memory (LSTM) networks, a Deep Learning (DL) technique, with Shapely Additive Explanations (SHAP), an Explainable Artificial Intelligence (XAI) technique. This integration is expected to improve predictive accuracy and model explainability. Leveraging the strengths of LSTM in capturing complex sequential patterns in financial time series big data, the model incorporates technical indicators to enhance its performance in forecasting stock movements. Deep learning is known to have a "black box" nature, so incorporating XAI techniques aims to offer detailed insights into how input features contribute to model outputs. This integration of XAI enhances the interpretability of predictions and enables users to understand the underlying rationale of the model, fostering greater trust among investors and financial professionals. The study utilized stock price data from the Yahoo Finance Website. The model processed forecasting of Google stock prices. The practical utility of this approach is demonstrated through the decision-making module, which provides actionable buy, sell, or hold recommendations, showcasing its potential in real-world investment scenarios. Our results indicate a balanced synergy between prediction accuracy and explainability, establishing a transparent and reliable AI-driven financial forecasting framework. [ABSTRACT FROM AUTHOR]