Accurate Stock Price Forecasting via Feature Engineering and LightGBM

Abstract

 Stock price forecasting is a critical task in financial markets, as it allows investors to make informed decisions based on anticipated market trends. With the advent of machine learning, traditional methods of predicting stock prices have been complemented and often surpassed by data-driven techniques. Among these techniques, LightGBM (Light Gradient Boosting Machine) has shown remarkable performance due to its efficiency and scalability, especially when combined with feature engineering. This paper explores the role of feature engineering in improving the predictive accuracy of stock price forecasting using LightGBM. Various technical, statistical, and market-related features are generated and evaluated, followed by training a LightGBM model on these features. The findings suggest that feature engineering plays a vital role in enhancing the performance of LightGBM, leading to more accurate predictions in stock price movements. This research emphasizes the importance of domain knowledge in selecting relevant features and highlights the potential of machine learning methods in financial forecasting.