Air Quality Index Prediction Using Support Vector Regression Based on African Buffalo Optimization

Support Vector Regression (SVR) is one of the machine learning models widely used in regression analysis. As an alternative for fitting a line to the data points like typical linear regression algorithms, it finds a hyperplane that is effective in fitting data points in a continuous space. The kernel type and hyperparameters significantly influence the performance and effectiveness of SVR. Determination of the optimal values is crucial in ensuring the success of prediction, regardless of the application domain. This study adapts the African Buffalo Optimization (ABO) algorithm to determine SVR’s regularization and kernel parameters. The ABO algorithm mirrors African buffaloes’ hunting and defensive behavior, offering ability to track the best position and extensive memory capacity to discover the best solution for problems under analysis. Evaluation is then performed on the air quality index benchmark dataset, and the prediction results of SVR-ABO are compared against other optimized SVR prediction models. The results show that SVR-ABO is a better algorithm because it produces smaller errors and best fits the data. Such an outcome indicates that the proposed SVR optimized by ABO is a competitive prediction model in data analytics.