EMPIRICAL ANALYSIS: SVR WITH DIVERSE KERNEL MODELS FOR FORECASTING AMAZON STOCK PRICE

Abstract

Current stock price prediction methods often struggle with non-linear market behavior. This paper addresses this limitation by proposing a Support Vector Regression (SVR) approach with multiple optimized kernel functions for improved Amazon stock price prediction from 1997 to 2023. This research explores the effectiveness of four kernels: wavelet, Ensemble Hyperbolic Tangents with Bayesian Optimization (EHT-BO), Combined Polynomial Radial Basis Function (RBF) with Grid Search (CPR-GS), and Hybrid ANOVA Polynomial RBF with Grid Search (HAPR-GS). The models are evaluated using various regression metrics. Among them, the CPR-GS kernel SVR model achieves the best performance with a Mean Absolute Error (MAE) of 0.159, indicating its potential for accurate stock price prediction.