Fuzzy Feature Engineering Integrated Artificial Neural Network for Stock Market Index Prediction

The prediction of stock market index prices is a critical and complex task, influenced by nonlinear and dynamic factors such as market trends and volatility. Recent advancements in Machine Learning (ML) and Deep Learning (DL) techniques, particularly Artificial Neural Networks (ANNs), have significantly enhanced predictive capabilities. These models effectively capture intricate relationships in large datasets, yet their performance is often limited by noisy and unstructured input features. This study introduces a novel approach that combines fuzzy feature engineering with ANN to predict stock market index prices. The methodology involves preprocessing historical stock data, including High, Low, Open, and Close (OHLC), and generating technical indicators like moving averages and standard deviations. These features are fuzzified into linguistic categories (low, medium, high) using triangular membership functions, improving interpretability and reducing noise. The fuzzy features are then fed into an ANN model with multiple hidden layers to predict index close prices. The results demonstrate high accuracy across major indices, such as the Nifty 50 and BSE Sensex. For example, the BSE Sensex achieved an R2 value of 0.998 and a Mean Squared Error (MSE) of 0.004. Additionally, the learning curve showed rapid convergence within 50 epochs, reducing the loss from 0.4 to below 0.05, emphasizing the efficiency of fuzzy-engineered features. By integrating fuzzy logic with neural networks, this approach enhances prediction accuracy and interpretability, providing a robust tool for addressing the complexities of stock market forecasting.