Comparison of the Application of Linear Regression with Sliding Window Validation and K-Fold Cross-Validation for Forecasting Covid-19 Recovered Cases

Abstract

The increase in confirmed cases and deaths due to Covid-10 continues to spread and increase day by day throughout the world. This has resulted in a world health crisis that impacts all sectors of life. The government declared a movement to suppress the spread of Covid-19, so it is necessary to understand the pattern of Covid-19 problems. Researchers contribute scientifically to finding patterns of death or recovery due to COVID-19 by applying Machine Learning methods. The Linear Regression and Sliding Window preprocessing methods are appropriate for forecasting time series data. This research obtained RMSE results at 0.320 with linear regression with sliding window validation and RMSE at 0.320 with linear regression with K-Fold cross-validation. This proves that Linear Regression with Sliding Window Validation can improve performance much better than k-fold cross-validation in forecasting COVID-19 recovery cases in China. The sliding window validation method has been proven to increase accuracy for forecasting with time series data compared to other standard preprocessing methods, namely K-Fold cross-validation. In the future, further research is needed to test different types of time series data by comparing the application of sliding window validation and K-Fold cross-validation or developing other validation models.

Keywords

• covid-19
• forecasting
• linear regression
• sliding window validation