Stock Price Prediction Using Feature Engineering and LightGBM

Abstract

Stock price prediction is a critical area of research in financial markets, driven by the need for robust, data-driven methods to guide investment decisions. Traditional models often struggle to capture the complex, nonlinear relationships in financial data. This paper explores the application of feature engineering and Light Gradient Boosting Machine (LightGBM) to enhance predictive accuracy. Feature engineering is employed to extract meaningful features from raw stock data, incorporating technical indicators, temporal patterns, and sentiment analysis. LightGBM, known for its efficiency and scalability, is used to handle large datasets and capture intricate relationships in the data. The proposed approach is tested on historical stock market data, demonstrating superior performance compared to conventional machine learning methods. The results emphasize the importance of carefully crafted features and advanced ensemble methods for predicting stock prices.