Work

Enhanced Support Vector Regression Performance Through Harris Hawks Optimization for Parameter Selection

Year: 2024
Type: article
Abstract: Support Vector Regression (SVR) is a powerful machine learning technique widely applied in time series forecasting and various prediction tasks. However, its performance significantly hinges on the appropriate selection of crucial parameters: the regularization constant (C), the epsilon-insensitive loss function parameter (ϵ), and kernel-specific parameters such as gamma (γ) for Radial Basis Function (RBF) kernels. Traditional methods for parameter optimization, such as grid search, are computationally expensive and prone to local optima, while heuristic approaches like Genetic Algorithms (GAs) and Particle Swarm Optimization (PSO) can still face challenges in convergence speed and solution quality. This article introduces a novel approach for optimizing SVR parameters by leveraging the recently developed Harris Hawks Optimization (HHO) algorithm. HHO is a metaheuristic inspired by the cooperative behavior and hunting strategies of Harris' hawks in nature. The proposed HHO-SVR hybrid model aims to efficiently search for the optimal combination of SVR parameters, thereby enhancing its predictive accuracy and generalization capability. This paper details the theoretical foundations of SVR and HHO, the methodology for their integration, and hypothetical experimental results demonstrating the effectiveness of the HHO-SVR model compared to other established optimization techniques in improving forecasting performance metrics such as Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE). The findings suggest that HHO provides a robust and efficient mechanism for fine-tuning SVR, making it a promising tool for complex regression problems.
Language: English
Open Access status: diamond
OpenAlex ID: W7220117279
FWCI:
Cites:
Cited by:
Related to: 10