Optimization of Support Vector Regression for Improved Prediction Accuracy in Cross-Sectional Data Using Genetic Algorithms

Abstract

Support Vector Regression (SVR) is a machine learning technique designed to predict continuous values by extending the principles of Support Vector Machines (SVM) into regression tasks. The performance of SVR models can be constrained by the selection of hyperparameters, which significantly affect the model’s predictive accuracy. To overcome this challenge, Genetic Algorithms (GA) can be utilized to optimize the hyperparameters of the SVR model. The GA demonstrated a steady improvement in fitness over 100 iterations. In this study, researchers focus on optimizing SVR for improved predictive accuracy in analysing cross-sectional data related to COVID-19 pandemic in Sulaymaniyah governorate. By leveraging GA for hyperparameter tuning, our research aims to evaluate the performance of a SVR with GA combined for optimizing complex, non-linear relationships in cross-sectional data, and improve the accuracy of the SVR model through GA. While previous research has explored optimizing similar models, to the best of the researchers' knowledge, this is the first study to apply such an optimized model to this specific dataset in Iraq and for medical field. The integration of SVR with Genetic Algorithms represents a novel approach in predictive modeling for COVID-19 pandemic related complications. Initially, the GA achieved a low mean fitness value of 0.0451, which steadily increased, reaching a peak of 0.0792. The results underscore the efficacy of this hybrid approach in finding optimal solutions, with predictions showing good alignment with actual data values. Overall, the integration of GA and SVR provided a robust method for solving complex optimization problems.