APPLICATION OF GRADIENT BOOSTING FOR STOCK PRICE PREDICTION

Solobuto Aleksei Viktorovich, Pavlov Valeriy Anatolyevich

This study examines the application of gradient boosting as a tool for forecasting the value of securities. This machine learning method has proven to be one of the most effective algorithms for solving regression and classification problems due to its ability to model complex nonlinear relationships and its robustness against overfitting when parameters are properly tuned. To improve prediction accuracy, a careful selection of informative financial and technical indicators that potentially influence the value of securities was conducted. Feature selection was performed based on both expert evaluations and statistical methods for analyzing variable significance. Additionally, the Parzen window method [1], a type of Bayesian optimization, was used for automated parameter tuning. This approach enables efficient exploration of parameter space.

Keywords:gradient boosting, forecasting, feature engineering, stocks, indicators.