Stocks Market Analysis and Prediction using Gradient Boosting Neural Network

Abstract

The financial sector is characterized by a highly complex, dynamic, and non-linear structure, which makes forecasting stock market performance a challenging task. Over the last decade, extensive research has focused on mining financial time series data using both data mining techniques and traditional statistical methods. By extracting technical indicators from historical prices and volumes, and transforming them into predictive models, technical analysis provides valuable insights into stock behavior. However, identifying highly informative financial indicators requires deep domain knowledge, which is not always readily available. This research aims to design a robust and effective framework for data normalization and feature extraction, addressing the challenges of stock market forecasting. The raw stock data is initially acquired and transformed into a synthetic format suitable for preprocessing. Feature selection techniques are then applied to identify the most relevant characteristics. Classification is carried out using a gradient boosting neural network approach, and the model’s predictive performance is evaluated using two widely adopted metrics: mean absolute error (MAE) and root mean square error (RMSE).