English Translation: Data Splitting and Model Evaluation in Time Series Analysis

To prevent potential time dependency, the data from January 2017 to December 2020 were used as both the training and verification sets for model training. 20% of the data were randomly selected as the verification set, while the remaining 80% were used as the training set. The data from the entire 21-year period were chosen as the testing set, not just randomly, but because we needed to evaluate the forecasting performance for a complete year, including daily, monthly, and seasonal variations. However, not all results were presented in the article.