Enhancing Stock Market Prediction Using Gradient Boosting Neural Network: A Hybrid Approach

Abstract

This paper introduces an innovative paradigm in cryptocurrency market analysis and prediction by exploiting the potency of the gradient boosting neural network (GBNN). This pioneering machine learning model amalgamates neural networks and gradient boosting techniques to offer a robust methodology. To enhance the GBNN's predictive capabilities, we enriched its input data with a spectrum of technical indicators. Moreover, we employed the support vector regressor for feature engineering, contributing to the exclusion of insignificant variables. We coined the term "hybrid approach" to describe our pipeline, employing it to train the GBNN model using historical cryptocurrency data. A multitude of experiments were conducted to demonstrate the superior performance of our approach in terms of model accuracy and error on previously unseen data. Notably, our proposed method outperformed state-of-the-art machine learning models, showcasing its effectiveness.