Times series data analysis: The Holt-Winters model for rainfall prediction In West Java

Abstract

Time series data analysis is used to analyze data that considers time and data characteristics to predict future events. One of the time series data is rainfall data. Rainfall data has a seasonal pattern because there is a pattern that repeats itself over a certain period. Data analysis that considers the characteristics of seasonal patterns is the Holt-Winters method. The Holt-Winters model is divided into two, namely additive and multiplicative models. This research aims to compare the Holt-Winters additive and multiplicative methods to see the accuracy in predicting rainfall data in West Java. The additive model has level parameter α=0,435, trend parameter β=0, seasonal parameter γ=1, and RMSE value 140,174. The multiplicative model has level parameter α=0,936, trend parameter β=0, seasonal parameter γ=0,247, and RMSE value 150,020. The additive model has a smaller RMSE value so it can predict future rainfall with greater accuracy.