Abstract

The post-COVID-19 outbreak and the second and third waves experienced in certain parts of the world remain a yardstick for predicting the likelihood of another pandemic. The attendant global health crisis has made accurate predictions an essential ingredient for effective management and response. With the attention given to artificial intelligence and machine learning techniques and the phenomenon of outcomes in multiple domains of healthcare, it is imperative to understand the underlying factors that make for efficient results. The challenge of handling the parameters during model design and implementation is a major motivation behind this study. The primary objective of this paper is to leverage the context of COVID-19 to explain the impact of parameter tuning on the prediction accuracy of machine learning algorithms. The performances of five different algorithms, namely: ordinary least squares (OLS), support vector with linear kernel (SVMLK), support vector with radial kernel (SVMRK), tuned support vector with linear kernel (SVMTLK), and tuned support vector with radial kernel, were compared.

The metrics mean absolute error (MAE), root-mean-square error, and normalized root-mean-square error are used to assess the performance of each of the algorithms. The comparison of the algorithms provides us with the best-performing algorithms for forecasting COVID-19 outcomes. OLS is very inefficient for modeling complex or nonlinear scenarios. SVMs are, however, capable of handling high-dimensional nonlinear data. SVMRK, unlike SVMLK, is comfortable capturing complex patterns through data transformation into a higher-dimensional feature space. However, parameter tuning unlocks the full potential of SVMs. This paves the way for SVMTLK, which optimizes the hyperparameters of the algorithms.

The outcome of this research indicates high performance by tuned support vector regression with a radial kernel, outperforming all other techniques. With the least error in the prediction, the five COVID-19 outcome variables. The importance of proper parameter tuning in machine learning algorithms is clearly highlighted, and the emphasis on ensuring careful tuning is highlighted as a key step in developing reliable machine learning models. These findings are very important to researchers and practitioners, especially in the health field. Understanding the impact of parameter tuning is therefore a useful issue when making informed decisions and when applying machine learning techniques to complex real-world problems.