%)  These findings highlight that national fisheries management policies and environmental variables have a significant influence on catch production, underscoring the importance of data-driven approaches in achieving sustainable fisheries management. This study also emphasizes the need to integrate machine learning techniques with conventional fisheries management strategies to enhance prediction accuracy and strengthen the foundation for policy decision making in Indonesia’s tropical marine ecosystems."> %)  These findings highlight that national fisheries management policies and environmental variables have a significant influence on catch production, underscoring the importance of data-driven approaches in achieving sustainable fisheries management. This study also emphasizes the need to integrate machine learning techniques with conventional fisheries management strategies to enhance prediction accuracy and strengthen the foundation for policy decision making in Indonesia’s tropical marine ecosystems.">
Laporkan Masalah

Pemodelan Data Hasil Tangkapan Purse Seine Pelagis Besar Periode 2015-2024 di WPPNRI 572 dan 573 Menggunakan Random Forest dan ARIMA

Nadya Rahadini, Dr.rer.nat. Riza Yuliratno Setiawan, S.Kel., M.Sc. ; Prof. Dr. Ir. Djumanto, M.Sc.

2025 | Tesis | S2 ILMU PERIKANAN

Sektor perikanan tangkap merupakan komponen penting bagi pembangunan ekonomi dan ketahanan pangan Indonesia, khususnya di Wilayah Pengelolaan Perikanan (WPPNRI) 572 dan 573 yang memiliki peran strategis bagi spesies pelagis besar. Penelitian ini mengevaluasi kinerja peramalan dari dua model, yaitu AutoRegressive Integrated Moving Average (ARIMA) dan Random Forest (RF), dalam memprediksi data tangkapan bulanan kapal purse seine pelagis besar di wilayah tersebut selama periode 2015 hingga 2024. Studi ini menggunakan data tangkapan harian untuk spesies cakalang, tuna mata besar, tuna sirip kuning, layang, dan tongkol. Prosedur penelitian mencakup proses standarisasi data, penerapan imputasi nol untuk mengatasi nilai yang hilang, serta pengagregasian data ke dalam interval bulanan. Model dievaluasi menggunakan metrik Root Mean Square Error (RMSE) dan normalized RMSE (NRMSE) guna menilai tingkat akurasinya dalam menggambarkan variabilitas tangkapan. Hasil penelitian menunjukkan bahwa model RF memberikan kinerja yang lebih unggul dibandingkan ARIMA, dengan nilai RMSE yang lebih rendah serta kemampuan adaptasi yang lebih baik terhadap karakteristik data yang bersifat nonlinier dan nonstasioner, dengan kesalahan prediksi relatif tinggi (RMSE/Mean >30%). Sedangkan RF memberikan hasil yang lebih akurat (RMSE/Mean <20>machine learning dengan strategi pengelolaan perikanan konvensional untuk meningkatkan akurasi prediksi dan memperkuat dasar pengambilan kebijakan pada ekosistem laut tropis Indonesia.

 

The capture fisheries sector is a vital component of Indonesia’s economic development and food security, particularly within Fisheries Management Areas (FMA) 572 and 573, which play a strategic role in supporting large pelagic species. This study evaluates the forecasting performance of two models AutoRegressive Integrated Moving Average (ARIMA) and Random Forest (RF) in predicting monthly catch data of large pelagic purse seine vessels in these regions for the period 2015–2024. The study utilizes daily catch records for skipjack tuna, bigeye tuna, yellowfin tuna, scads, and bullet tuna. The research procedures include data standardization, zero-value imputation to address missing values, and aggregation of the dataset into monthly intervals. Model performance was assessed using the Root Mean Square Error (RMSE) and normalized RMSE (NRMSE) metrics to evaluate their accuracy in representing catch variability. The results show that the RF model outperformed ARIMA, yielding lower RMSE values and demonstrating greater adaptability to the nonlinear and non-stationary characteristics of the data, although overall prediction errors remained relatively high (RMSE/Mean > 30%). In contrast, RF achieved more accurate predictions (RMSE/Mean < 20 lang="id">%)  These findings highlight that national fisheries management policies and environmental variables have a significant influence on catch production, underscoring the importance of data-driven approaches in achieving sustainable fisheries management. This study also emphasizes the need to integrate machine learning techniques with conventional fisheries management strategies to enhance prediction accuracy and strengthen the foundation for policy decision making in Indonesia’s tropical marine ecosystems.

Kata Kunci : Pelagis besar, Purse seine ARIMA, Random Forest, Samudra hindia.

  1. S2-2025-527224-abstract.pdf  
  2. S2-2025-527224-bibliography.pdf  
  3. S2-2025-527224-tableofcontent.pdf  
  4. S2-2025-527224-title.pdf