Deep Learning based Stock Trading Strategies using Leading Multi-Indicator Confirmations

This study presents a novel Reinforcement Learning (RL)-based stock trading strategy that integrates the Simple Moving Average (SMA) crossover with confirmation signals from Average Traded Volume and the Money Flow Indicator (MFI). The framework proposed implements a Q-learning algorithm to develop a policy that dynamically adapts to market conditions while maximizing returns while minimizing risk. SMA crossover is a classic trend-following technique, provides primary trade signals, while volume and MFI confirmations strengthens decision accuracy by validating price momentum and capital flow strength. Historical daily data from selected stocks in the Indian market spanning five years was used for training and back testing. The RL agent was trained using a reward function calibrated on cumulative returns, Sharpe ratio, and drawdown metrics. Comparative results against traditional rule-based strategies and machine learning classifiers indicated that the RL-based model outperformed in terms of profitability and risk-adjusted returns. The inclusion of volume and MFI confirmations significantly reduced false signals and overfitting issues typically encountered in volatile market phases. This hybridized approach highlights the potential of merging technical indicators within RL frameworks for intelligent, autonomous trading systems. The study contributes to the evolving field of financial machine learning by offering a robust, data-driven decision-making model for enhanced stock market performance.