An AI-Based Stock Market Trading Strategy Advisor Integrating LSTM Prediction, FinBERT Sentiment Analysis and Deep Q-Network (DQN) Reinforcement Learning

Abstract: Financial markets can be influenced by quantitative price movements or by qualitative behavioral factors such as news sentiment and investor psychology. Historical stock prediction models are traditionally based on data, but they are usually not good for finding any trading decisions and tend to fail us. This paper presents an AI stock market trading strategy advisor system that combines Long Short-Term Memory (LSTM) networks (for price prediction) with FinBERT-based Natural Language Processing (NLP) using financial sentiment and Deep Q-Network (DQN) reinforcement learning as deep neural networks for intelligent decision making. It simulates actual trading conditions using historical market data and financial news. The LSTM module predicts future price trends, FinBERT estimates sentiment based on financial headlines, and the DQN agent learns with regard to the return and sentiment signal the best trading decisions including Buy / Sell / Hold in the market based on the expected values and direction of the returns. Experimental study shows that the integrated strategy of this approach in their experimental implementation enables better decision-making support than the classic strategies based on combining technical forecasting with behavior-based decision support, because their combination is more effective than the application-level forecasting approach.

Keywords: AI Trading Advisor, Deep Q-Network (DQN), FinBERT, LSTM, Reinforcement Learning, Sentiment Analysis, Stock Market Prediction