<![CDATA[Abstrak – Penelitian ini mengembangkan sistem prediksi harga Bitcoin menggunakan metode ARIMA dan SARIMA dengan data real-time dari API CoinGecko. Data di kumpulkan dalam 4 interval waktu (15 menit, 30 menit, 1 jam, 1hari) dan dibagi menjadi 80% data training serta 20% data testing. Uji stasioneritas Augmented Dickey-Fuller (ADF) awal menunjukan data tidak stasioner (p-value=0,5725), namun menjadi stasioner setelah di proses differencing (p-value=0,0000). Hasil evaluasi model menunjukan bahwa SARIMA secara signifikan lebih unggul di bandingkan ARIMA. Model SARIMA menghasilkan MAPE 0.76%, MAE 843,27, RMSE1.095,43, dan korelasi 0,92. Sementara itu, ARIMA hanya menghasilkan MAPE 2.81%, MAE 3.091,65, RMSE 3.750,52, dan korelasi -0,28. Keunggulan SARIMA di sebebkan kemampuan menangkap pola musiman dalam data. Sistem ini berhasil diimplementasikan menggunakan framework Streamlit dengan fitur auto-refresh 30 detik dan visualisasi candlestick chart interaktif. Sistem prediksi real-time ini dapat menjadi alat pendukung keputusan investasi yang akurat bagi investor cryptocurrency.Kata kunci : Prediksi Bitcoin; ARIMA;SARIMA; Stremalit; API CoinGecko; Abstract - This research develops a Bitcoin price prediction system using ARIMA and SARIMA methods with real-time data from the CoinGecko API. Data was collected in 4 time intervals (15 minutes, 30 minutes, 1 hour, 1 day) and split into 80% training data and 20% testing data. The initial Augmented Dickey-Fuller (ADF) stationarity test showed the data was non-stasionary (p-value=0.5725), but became stasionary after the differencing processs (p-value=0.0000). Evaluation results show that SARIMA performed significantly better than ARIMA. The SARIMA model yielded a MAPE of 0.76%, MAE of 843.27, RMSE of 1.095,43, and a correlation of 0,92. In contrast, the ARIMA model only achived a MAPE of 2,81%, MAE of 3.091,56, RMSE 3.750,52, and a correlation of -0,28. SARIMA’s superiority is attributed to its ability to capture seasonal patterns in the data. The system was successfully implemented using the streamlit framework, featuring 30-second auto-refresh and interactive candlestick chart visualization. This real-time prediction system can serve as an accurate decision support tool for cryptocurrency investors.Keywords: Prediction Bitcoin; ARIMAt; SARIMA; Streamlit; API CoinGecko.]]>
