Prediction of COVID-19 Cases in Malaysia by Using Machine Learning: A Preliminary Testing

Estimating the spread of disease like COVID-19 is a crucial task that can reduce its impact and help the authorities to respond. Currently, the number of active cases in Malaysia is less than 50% of the accumulated points. This paper will present the most accurate machine learning model to predict the number of new possibilities for 01/04-15/04/2020 (15 days). Notably, four machine learning models and one traditional statistical prediction analysis ARIMA have been trained, tested and evaluated. The results are compared to the actual data. We found that the multilayer perceptron neural network has outperformed all models in estimating the number of new positive cases in a specific time range. However, the ARIMA model performs better in predicting recovered and death cases than the machine learning models. We have applied the root mean square error (RMSE) and mean absolute predictive error (MAPE) evaluation metrics to assess the model performance. The proposed approach can be viewed as a feasible alternative to forecast the new COVID-19 cases in Malaysia, where there is no constant seasonal trend and no past data sources for comparison. The proposed methodology is easily tractable and computationally efficient to demonstrate the number of new cases that are useful and beneficial in terms of control management.