Application of Time Series Model in Relative Humidity Prediction

Abstract

Relative humidity is the percentage of the vapor pressure in the air and the saturated vapor pressure at the same temperature. In the context of global climate change, accurate and reliable relative humidity prediction is of great significance in all fields. In this paper, taking the relative humidity data of Ya'an city, Sichuan province as an example, we used three methods, Holt-Winters, SARIMA and XGBoost to establish a time series model to predict relative humidity. We use the grid search and other methods for selecting the optimal parameters, the mean absolute error, mean square error and mean absolute percentage error as evaluation standard, the experimental results show that although the three model error is within an acceptable range, but XGBoost model predicted results are more accurate, better performance and stronger ability to resist a fitting, obviously better than the other two models, which can provide a reference for practical work.