Jakarta, INTI - The Research Center for Data Science and Information (PRSDI) of the National Research and Innovation Agency (BRIN) has developed PrediaBeat, an artificial intelligence (AI)-based healthy food recommendation system.
This system is designed to help individuals with prediabetes choose foods that meet their health and nutritional needs. This innovation was developed by Kokoy Siti Komariah, a researcher at PRSDI BRIN, and her team.
PrediaBeat was developed due to the increasing number of people with prediabetes, both in Indonesia and globally. Prediabetes often has no symptoms, so many individuals are unaware that their blood sugar levels are above normal limits. However, without treatment through dietary and lifestyle changes, this condition carries the risk of developing into type 2 diabetes mellitus.
Assisting with Personalized Food Recommendation
Kokoy explained that the PrediaBeat system was developed to address the public's need for more personalized, easily accessible, and evidence-based nutritional information. Thus, this system is expected to help users make more informed decisions regarding their daily food consumption.
"Information about healthy eating is currently scattered across various sources and is often general in nature. Through PrediaBeat, we strive to provide personalized food recommendations based on each user's health condition, preferences, and nutritional needs," she said.
Unlike conventional health chatbots, PrediaBeat integrates a Large Language Model (LLM) with a knowledge base containing nutritional data, glycemic index, nutrient content, and prediabetes management guidelines supported by scientific references.
PrediaBeat is also designed using a User-Centered Design (UCD) approach, so the interface and user experience are developed based on the needs of potential users. Kokoy hopes this approach will improve ease of use and technology acceptance by the public.
How PrediaBeat is Developed
In its operation, PrediaBeat uses Retrieval-Augmented Generation (RAG) technology, which combines the capabilities of a Large Language Model with a trusted knowledge database. This technology enables the system to generate more accurate and relevant recommendations and information, while minimizing the risk of inaccurate information.
Kokoy stated that the development of PrediaBeat was carried out in a multidisciplinary manner, involving researchers in the fields of artificial intelligence, natural language processing, food computing, nutrition, and public health. This research also opens up opportunities for collaboration with universities, hospitals, and healthcare institutions to support data development, scientific validation, and field implementation.
"PrediaBeat has now entered the prototype development stage. The research team has completed user needs analysis and system design at the Minimum Viable Product (MVP) level. This year's development focus includes building a nutritional knowledge-based food database and developing an AI-based chatbot," explained Kokoy.
BRIN will continue refining the recommendation model, testing system performance, and evaluating user experience. PrediaBeat is expected to evolve into a digital health assistant that supports diabetes prevention through healthy food recommendations, personalized education, monitoring consumption patterns, and integration with fitness apps and wearable devices.
"Through PrediaBeat, we hope that AI technology can be utilized responsibly to support the transformation of digital healthcare services in Indonesia while helping the public adopt healthier and more sustainable lifestyles," said Kokoy.
Conclusion
BRIN, through the Data and Information Science Research Center (PRSDI), developed PrediaBeat, an AI-based healthy food recommendation system to help people with prediabetes choose foods according to their health condition and nutritional needs. This system provides personalized recommendations based on user profiles, such as age, gender, health conditions, food preferences, and nutritional needs, and was developed using a User-Centered Design (UCD) approach. Currently, PrediaBeat is still in the prototype and Minimum Viable Product (MVP) development stage, with a focus on building a nutritional database and AI chatbot.
Read more: BSSN Encourages Hospitals to Strengthen Cybersecurity in the Era of Digital Transformation