The application of computational methods like Machine Learning (ML) and Artificial Intelligence (AI) in several fields has recently received increased attention from academicians and field professionals. However, the availability of several of these computational techniques requires expertise to make the right choice. In making things easy for non-experts and inexperienced practitioners to make a suitable choice, some studies sought to compare the performance of different ML algorithms (MLAs) in solving problems in unique fields. However, most of the previous studies were domain centered on a handful of MLAs, which does not give a fairground for all-inclusive comparative analysis. Also, many of these studies compared MLAs for classification tasks to the best of our knowledge without considering the same MLAs performance for regression problems. This paper examines the performance of twenty-one (21) MLAs for classification and regression tasks based on six datasets from different domains. Empirically we compare their prediction results based on accuracy, balanced accuracy, F-score, Area Under the Curve (AUC), root mean square error, r-squared and adjusted r-squared. The random forest algorithm gave a consistent performance across all the six datasets in classification and regression tasks. However, on average, the XGBoost outperformed all MLAs applied in this study. The dummy algorithm and the linear regression were more moderate than the rest of the applied MLAs in computational complexity. Nevertheless, the overall study outcome shows that MLAs algorithms efficiently solve everyday challenges in elections outcome, financial fraud, network intrusion detection, meteorological forecast and heart diseases discovery; but their performance varies across domains and dataset dimensions.