Application of Clustering and VARIMA for Rainfall Prediction

Abstract

Abstract Global warming causes climate change throughout the world which has an impact on changes in erratic rainfall. For fishermen, the estimated rainfall is an important thing to know before sailing. VARIMA is divided into VAR, VMA, VARMA, and VARIMA. This study uses a rainfall dataset consisting of rainfall (mm), temperature (°C), humidity (%) and wind speed (km/hour) with observations per day for one year in 2018. Training data is data that starts from 1 January 2018 until 30 November 2018 and the testing data starts on 1 December 2018 until 31 December 2018. Clustering was carried out in the area around Lampulo Pelabuhan Perikanan Samudera (PPS) which are Bandar Raya, Lampineung, Kuta Radja, and Kuta Alam. The best model obtained is VARIMA (1,1,2), which is based on the smallest AIC and BIC values, which are 8.5104 and 9.0581. The VARIMA (1,1,2) model also fulfills the white noise test, so it can be concluded that the data does not have autocorrelation and model can be used for predictions. Estimated values of rainfall, temperature, humidity and wind speed are not much different from actual values, with MAPE of all variables are relatively low.