Over the last year, the COVID-19 epidemic has afflicted over 150 million individuals and killed over three million people globally. Various forecasting models attempted to estimate the temporal course of the COVID-19 pandemic during this time period in order to determine effectiveness of the government action in facing COVID-19 outbreak. In this study, Autoregressive Integrated Moving Average (ARIMA) models were used in order to forecast the COVID-19 mortality rates data in Malaysia. The accuracy of the ARIMA models is then evaluated by using Mean Absolute Error (MAE) and Root Mean Square Absolute Error (RMSE). The forecasting model with the lowest error is picked as the best. In this study, ARIMA (1,1,3) outperformed the ARIMA (1,1,2) and ARIMA (1,1,4) models since it has the lowest MAE and RMSE values. However, as compared to ARIMA (1,1,4), the study found that ARIMA (1,1,3) model is not adequate in terms of model fitting due to the errors were not normally distributed. Hence, ARIMA (1,1,4) model was chosen to make prediction of COVID-19 mortality rates. Accordingly, the findings through this study can be used as a preliminary study to predict the COVID-19 mortality rates and other future pandemic cases to mitigate risk of increasing cases.