PERAMALAN ALOKASI BBM SUBSIDI KERETA API DENGAN METODE STATISTIKA DAN MACHINE LEARNING

Abstract

Based on the results of the on-desk verification on June 2nd, 2021, it was found that there was a quite significant overstock of JBT refined fuel oil for PT KAI (Indonesian Railways Company) during the Covid-19 pandemic, namely 42.9% in the first quarter of 2021 for passenger train types. The difference of 42.9% came from the quota for passenger train fuel for the first quarter of 37,129 kl compared to the realization of passenger train fuel for the first quarter of 15,753 kl. Therefore, the purpose of this study is to find out forecasting with statistical methods and machine learning in solving the overstock problem. This study uses forecasting with statistical methods, namely cyclical, cyclical, exponential, quadratic, and linear regression methods; and machine learning methods, namely ARIMA (AutoRegressive Integrated Moving Average) to find the best JBT (certain types of fuel) refined fuel oil allocation scenario for PT KAI with the benchmark is the measurement that yields the smallest error using the MAPE (Mean Absolute Percentage Error) method. The method with the smallest error is the exponential method with a MAPE value of 7.37% with the determination of the JBT refined fuel oil allocation for the passenger train section of PT KAI of 8,474.52 kiloliters for the 3rd quarter of 2021, 7,836.58 kiloliters for the 4th quarter of 2021, 7,246.65 kiloliters for the 1st quarter of 2022, and 6,701.14 kiloliters for the 2nd quarter of 2022. Overall, BPH MIGAS (Oil and Gas Downstream Regulatory Agency) will set 30,258.90 kiloliters for the allocation of JBT refined fuel oil for the passenger trains of PT KAI for the next 4 quarters (12 months). The forecast only considers 1 variable, namely historical data on the use of PT KAI JBT refined fuel oil for passenger trains. This research was conducted because there was no previous research that forecast the refined fuel oil needs of the passenger train section of PT KAI that was heavily affected by Covid-19, especially with statistical methods and machine learning. The results of this study are recommended to BPH MIGAS as a consideration in determining the allocation of PT KAI's JBT refined fuel oil in the future during the Covid-19 pandemic.