Perbandingan Metode Imputasi Data Hilang untuk Peramalan Curah Hujan dengan ARIMA di Palangka Raya

Abstract

Rainfall is one of the fundamental factors influencing various aspects of a region, including the city of Palangka Raya. Therefore, statistical tools are needed to forecast rainfall for mitigation and planning related to its effects. The objective of this study is to evaluate the performance of the ARIMA model in forecasting rainfall and to analyze the impact of three methods of missing data imputation on forecast accuracy. Therefore, this study proposes an evaluation of the autoregressive integrated moving average (ARIMA) model's performance for rainfall forecasting and the handling of missing data through three imputation methods: mean imputation, forward filling, and backward filling. The data used in this study consists of historical rainfall data from 2013 to 2022. The forecasting results indicate that rainfall data modelled using ARIMA and treated for missing data via mean imputation demonstrated the best forecasting accuracy in the validation test, as rainfall data in Palangka Raya City is fluctuating without a consistent seasonal pattern. Thus, the historical average value is more representative than the preceding or subsequent values, which could reinforce local bias. This is demonstrated by the smallest RMSE, MAE, and MAPE values, which are 124,17%, 103,57%, and 40,32%, respectively.