Detection of Nutritional Status using K-Nearest Neighbors on a Mobile Based Platform

In recent decades, health issues related to nutritional status have become a major concern for the Indonesian government. Malnutrition or overnutrition can severely impact individual health, especially in children and adolescents. If left unaddressed, these issues can lead to various diseases, ranging from malnutrition to obesity and their associated complications. Despite the recognized importance of monitoring nutritional status, several challenges remain. Manual monitoring systems require significant time and resources. Moreover, access to healthcare services and qualified medical personnel for regular nutritional assessments is limited, particularly in remote or underserved areas. Indonesia’s geographical complexity, consisting of thousands of islands, further complicates the equitable distribution of healthcare services. As a result, many cases of malnutrition or overnutrition go undetected early, causing delayed interventions. This research proposes the development of a K-NN-based mobile application to detect nutritional status. The application provides an initial diagnosis based on the user's physical parameters, such as weight, height, age, and gender. The dataset includes 120,999 samples, with 70% used for training and 30% for testing. Implementation of K-NN with k=7 achieved an accuracy of 91% on the test data, with the best performance in the normal category (F1-score 0.950), followed by stunted (0.889) and severely stunted (0.863). This platform has the potential to contribute to sustainable health systems, particularly in low-resource settings, by reducing reliance on energy-intensive infrastructure and minimizing the need for long-distance travel for healthcare. It could also support public health initiatives by enabling efficient large-scale population monitoring and reducing the environmental impact of traditional health services.