Kebijakan Berbasis Data: Analisis dan Prediksi Penyebaran COVID-19 di Jakarta dengan Metode Autoregressive Integrated Moving Average (ARIMA)

Abstract

Data dan informasi merupakan bagian penting dalam pertimbangan mengambil keputusan terkait penanganan COVID-19. Data COVID-19 baik demografi maupun agregat di Provinsi DKI Jakarta diolah dan dianalisis untuk memberikan informasi mengenai situasi dan kondisi terkini terkait pandemi COVID-19 di Provinsi DKI Jakarta. Data COVID-19 tersebut juga dimanfaatkan untuk analisis prediktif untuk mengetahui perkiraan jumlah kasus COVID-19 di masa depan. Analisis prediktif yang digunakan dalam artikel ini adalah metode Autoregressive Integrated Moving Average (ARIMA). Model ARIMA merupakan salah satu metode forecasting hasil dari perluasan model Autoregressive Moving Average (ARMA) untuk data yang tidakstasioner. Analisis dan visualisasi data dilakukan menggunakan program Python dan Tableau dimana hasil analisis prediktif memperlihatkan tren kasus positif harian yang cenderung naik di kurun waktu 14 hari ke depan dari data yang digunakan. Hasil analisis ini dapat digunakan sebagai pertimbangan bagi pemerintah dalam mengambil kebijakan dan intervensi dalam penanganan COVID-19 di Jakarta, dan untuk masyarakat agar tetap melakukan tindakan preventif dalam mencegah kenaikan kasus, seperti mematuhi protokol kesehatan yang sudah ditetapkan oleh Pemerintah.

STL Decomposition and SARIMA Model: The Case for Estimating Value-at-Risk of Covid-19 Increment Rate in DKI Jakarta

“…Thus, we consider to apply Seasonal Autoregressive Moving Average (SARIMA) for modelling seasonal pattern in STL Decomposition. The SARIMA(0, 1, 2)(2, 1, 0) 12 shown a significant parameter and minimum AIC, 257.794. The VaR using Variance Covariance simulation by identifying seasonal, trend and noise based the linear combination STL-SARIMA framework which a MAPE results of 0.15 provides sharp and well prediction for extreme value with zero penalty.…”

“…SARIMA is a developing model from the previous model for data that does not contain seasonality, ARIMA [11]. From several research to forecasting COVID-19 cases using the ARIMA model have been carried out with good modeling accuracy results [12] [13] [14]. ArunKumar, et al used the ARIMA and SARIMA models for forecasting Covid-19 in 16 countries, then they found that the SARIMA model outperformed the ARIMA model in capture the seasonality or trends of the data [15].…”

“…It is necessary to select the order of SARIMA model, to choose order (𝑝, 𝑑, 𝑞)(P, D, Q) 𝑆 , we consider to identify based on the minimum AIC and statistical test of proposed parameters. The SARIMA (0, 1, 2)(2, 1, 0) 12 shown minimum AIC based Equation 4, 257.794. Moreover, the statistical test according to Equation 5, each parameters reveals a significant contribution to the SARIMA(0, 1, 2)(2, 1, 0) 12 model, the value of 𝑝 > |𝑧| of these parameters is less than 5%.…”

“…The SARIMA (0, 1, 2)(2, 1, 0) 12 shown minimum AIC based Equation 4, 257.794. Moreover, the statistical test according to Equation 5, each parameters reveals a significant contribution to the SARIMA(0, 1, 2)(2, 1, 0) 12 model, the value of 𝑝 > |𝑧| of these parameters is less than 5%.…”

“…According to the result of AIC, the optimal order of SARIMA model with the lowest AIC is (0, 1, 2)(2, 1, 0) 12 . Thus, 𝑆 𝑡 as seasonal of STL Decomposition can be written as follows,…”