Jakarta, INTI - The National Research and Innovation Agency (BRIN) is developing an artificial intelligence-based marine protein potential mapping tool to support the implementation of the Free Nutritious Meal Program (Makan Bergizi Gratis/MBG). This innovation is designed to ensure a secure marine protein supply based on precise, integrated, and sustainable spatial data.
The MBG program, as a national agenda, requires strong management, from targeting accuracy to logistical efficiency and accountability. The Head of BRIN's Electronics and Informatics Research Organization (OREI), Prof. Budi Prawara, said that data quality is key to the program's success.
"The program’s success is highly determined by data quality. This is where the role of data science, information, and digital governance becomes crucial," he said during the DESAIN #1 webinar, Wednesday, February 25, 2026.
BRIN Utilizes Deep Learning Technology to Map Fishing Zones
Currently, BRIN's Data Science and Information Research Center (PRSDI) is developing an AI-based model for mapping fishing zones for fish, shrimp, and seaweed to support the MBG program. This model integrates deep learning technology to generate spatially precise maps of species occurrence probabilities and estimates of marine protein potential.
Rezzy Eko Caraka, a PRSDI BRIN researcher and the originator of the marine protein map for the MBG program, explained that the research, titled "Kolmogorov–Arnold Network and ResNet for Marine Protein Mapping in Support of the Prabowo–Gibran MBG Program," aims to map Indonesia's marine protein potential as an alternative nutritional source for MBG food diversification.
The Supporting Data for The Mapping Tool Integrates Various Variables
Rezzy explained that the research was published in the reputable international journal, Discover Sustainability (Springer International Publishing), on October 22, 2025. The publication strengthened BRIN's academic contribution in supporting science-based policies.
Rezzy explained that the AI-based model integrates various oceanographic and water quality parameters, such as sea surface temperature, salinity, pH, dissolved oxygen, chlorophyll-a, water clarity, sea depth, and surface current velocity. Satellite data were obtained from various remote sensing platforms, including NASA through the MODIS instrument, SMAP, VIIRS, and OCO-2 instruments, as well as Landsat, SRTM, and OSCAR data.
Rezzy added that all of the variables were analyzed using the Kolmogorov-Arnold Network and Residual Neural Network (ResNet) approach to produce a map of marine protein potential in various Indonesian waters, from the Andaman Sea to the Arafura Sea.
He revealed that the test results showed a model accuracy rate of 94.6% in classifying marine protein potential. This system can identify potential zones for fish, various types of shrimp, and seaweed more efficiently and data-driven.
"This approach enables data-based fishing zone determination, making it more efficient and sustainable. We can also avoid overfishing through the concept of precision fishing and ecosystem-based management," Rezzy explained.
Going forward, Rezzy hopes that the results of this research will potentially be integrated into a national dashboard to support decision-making across ministries, agencies, and regional authorities. The integration of spatial data and AI is expected to provide a scientific foundation for strengthening national food security while optimizing Indonesia's marine protein potential for the sustainable success of the MBG Program.
Conclusion
BRIN develops an AI-based tool to map marine protein potential zones, which focuses on fish, shrimp, and seaweed. The mapping tool uses the Kolmogorov-Arnold Network and Residual Neural Network (ResNet) approach to analyze various variables and data, with results showing 94.6% accuracy in classifying marine protein potential.
Read more: OpenAI Secures $110 Billion in Funding, Reaching a $730 Billion Valuation