Stock Price Prediction Using Gradient Boosting Machine with Technical Indicators

This research endeavors to design a predictive model to forecast stock prices using a Gradient Boosting Machine (GBM) regressor, specifically emphasizing the adjusted closing price to mitigate the autocorrelation issues intrinsic to time series data. The investigation encompasses a 12-month duration of historical stock data, integrating two technical indicators, namely Exponential Moving Averages (EMA_5 and EMA_10), to augment predictive precision. Gradient Boosting, recognized for its proficiency in managing intricate and non-linear data patterns, was employed to forecast forthcoming stock prices. Critical performance indicators, including Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and the Coefficient of Determination (R²), played a role in evaluating model accuracy, attaining an R² score of 0.99, thereby signifying an almost perfect correlation between actual and predicted prices. Beyond numerical assessments, visual examinations through scatter plots, residual plots, and prediction error distribution further elucidated the model’s efficiency. The findings reveal that the amalgamation of Gradient Boosting with Technical Indicators such as EMA_5 and EMA_10, coupled with an emphasis on the adjusted closing price, proves exceptionally effective for stock price forecasting. This methodology achieved an accuracy rate nearing 99%, illustrating its substantial promise for stock market prediction and its proficiency in alleviating autocorrelation challenges within time series datasets.