<![CDATA[Produksi ikan hasil tangkapan yang didaratkan di Pelabuhan Perikanan Pantai (PPP) Bulu Tuban menunjukkan fluktuasi bulanan dengan pola musiman yang jelas, sehingga diperlukan pendekatan peramalan yang mampu merepresentasikan dinamika tersebut. Penelitian ini bertujuan mengidentifikasi pola deret waktu produksi ikan bulanan di PPP Bulu Tuban dan membangun model peramalan menggunakan Seasonal Autoregressive Integrated Moving Average (SARIMA). Data yang digunakan berupa data sekunder produksi ikan bulanan periode Januari 2019–Desember 2024 sebanyak 72 observasi. Pemodelan dilakukan mengikuti prosedur Box–Jenkins melalui visualisasi deret waktu, differencing non-musiman dan musiman , identifikasi kandidat model berdasarkan ACF dan PACF, serta estimasi dan seleksi model. Kinerja model dievaluasi menggunakan Bayesian Information Criterion (BIC), Root Mean Square Error (RMSE), dan Mean Absolute Percentage Error (MAPE), sedangkan diagnostik residual diuji dengan Ljung–Box. Hasil menunjukkan bahwa SARIMA merupakan model terbaik dengan BIC 10,797, RMSE 199,346, MAPE 207,607, dan residual yang memenuhi asumsi white noise . Model ini mampu merepresentasikan pola historis dan komponen musiman data, namun karena nilai MAPE masih tinggi, hasil peramalan lebih tepat diposisikan sebagai proyeksi pola umum produksi untuk tahun 2025 daripada prediksi numerik presisi tinggi pada setiap bulan. Fish catch production landed at Bulu Tuban Fishing Port exhibits monthly fluctuations with a clear seasonal pattern, indicating the need for a forecasting approach capable of representing such dynamics. This study aimed to identify the monthly time-series pattern of fish production at Bulu Tuban Fishing Port and to develop a forecasting model using the Seasonal Autoregressive Integrated Moving Average (SARIMA) approach. The analysis used secondary monthly fish production data from January 2019 to December 2024, comprising 72 observations. Modeling followed the Box–Jenkins procedure through time-series visualization, non-seasonal differencing, seasonal differencing, identification of candidate models based on ACF and PACF, and model estimation and selection. Model performance was evaluated using the Bayesian Information Criterion (BIC), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE), while residual adequacy was assessed using the Ljung–Box test. The results show that SARIMA was the best model, with a BIC of 10.797, RMSE of 199.346, MAPE of 207.607, and residuals satisfying the white-noise assumption . The model was able to represent the historical and seasonal structure of the data; however, given the relatively high MAPE, the forecast should be interpreted more cautiously as a projection of general production patterns for 2025 rather than a high-precision monthly numerical prediction.]]>
