AI-POWERED ANALYSIS

The study utilizes an XGBoost model for stock price prediction, chosen for its strong performance in time series forecasting and ability to capture complex data patterns. It employs a sliding window approach with a lookback period of 5 days to generate input-output sequences for training. Key technical features—Average Directional Index (ADX) and Stop and Reverse (SAR)—were added to the dataset to enhance accuracy. The data, comprising Nifty 50 stock prices from Jan 3rd, 2011, to July 31, 2024, underwent pre-processing, including removing missing values and outliers, feature selection, and normalization using z-score. The dataset was then split into 80% for training, 10% for validation, and 10% for testing. Hyperparameters for XGBoost were tuned to optimize performance and prevent overfitting, with settings like a max_depth of 5, n_round of 100, learning rate of 0.01, and gamma of 2.

The model's performance was evaluated using Mean Absolute Error (MAE), Mean Squared Error (MSE), and the Coefficient of Determination (R²). On the test set, the model achieved an average MAE of 0.00116, an average RMSE of 0.00156, and an impressive average R² of 85.310%. The maximum R² reached 90.669%, while the minimum was 67.565%. These results collectively demonstrate the model's strong predictive accuracy and its ability to capture a significant proportion of the price variability. The visualizations of predicted versus actual prices further confirm the model's capability to adapt to price variations. While performance is generally high, the variability in R² values across different data points indicates potential for further refinement and optimization in specific scenarios.

This research has significant implications for the finance industry. The AI-powered XGBoost model provides a robust tool for predicting expected returns and optimizing risk portfolios, moving beyond the limitations of traditional models. For financial analysts and portfolio managers, this means more effective diversification, enhanced risk mitigation, and the potential for higher returns. The study advocates for a shift towards integrating machine learning into standard investment strategies, promoting better-informed decisions and more resilient financial markets. However, the study also highlights the need for continued model refinement, addressing practical implementation challenges such as data access, regulatory compliance, and integrating explainability into AI models for greater transparency and trust.