Komparasi Peramalan ARIMA dan RNN pada Tunggakan Peserta BPJS Kesehatan Cabang Jambi

Abstract

BPJS Kesehatan is a public legal entity that is directly responsible to the President who has the task of providing Social Security to the community and has the authority to hold dues every month for participants. BPJS Kesehatan participants who are independent participants have an obligation to pay dues every month. Pandemic conditions affect participants' incomes in paying dues as a result participants' arrears rise. Arrears participants also occurred in BPJS Kesehatan Jambi Branch Office. BPJS Kesehatan branch office has SIMANIS (Sistem Penagihan Iuran Terintegrasi) application. Through this system, BPJS Kesehatan can reduce the arrears participants, especially Jambi branch offices. Thus, it makes easier for each branch office to evaluate the right policy, through arrears data it can be processed into forecasting information using traditional statistic methods ARIMA and machine learning methods RNN (Recurrent Neural Network). Therefore, this study aims to compare ARIMA and RNN in forecasting on premium arrears data of BPJS Kesehatan Jambi Branch Office participants. The dataset used is data on arrears of premium participants BPJS Kesehatan branch office Jambi per week. Through this forecasting, the results RMSE (Root Mean Square Error) and MAPE (Mean Absolute Percentage Error) from RNN 1 hidden layer 1 neuron method have RMSE  2739803819 and MAPE 1.97%, which is better than ARIMA (3,1,3) with RMSE 3238308628 and MAPE 2.62%. Then, to make it easier for non-programmer users to forecast built a graphical user interface (GUI) based on the web with Streamlit library. This research is expected to validate the forecasting model that is suitable for the arrears data of BPJS Kesehatan Jambi Branch Office participants.