Malnutrition among children in third-world countries like the Philippines remains a critical issue addressed by the UN’s Zero Hunger goal. Traditional methods such as K-Nearest Neighbor face limitations in accuracy due to bias. This study aims to enhance K-Nearest Neighbor’s predictive accuracy (KNN) by incorporating Gaussian kernel similarity and normalization techniques to predict malnutrition and BMI categories. Results show that Enhanced KNN consistently outperforms Basic KNN across various K values, achieving an average of 94.33% for bi-class (Malnutrition/Normal) and 92.67% for multi-class (BMI category) compared to Basic KNN's accuracy results: 87.43% for bi-class and 82.51% for multi-class. The Enhanced algorithm performs better across all metrics, averaging an improvement of 8%. Findings revealed that the use of Gaussian kernel enhanced the accuracy of K-Nearest Neighbor because it prioritizes nearer neighbors over distant ones, which reduces the influence of farther points. This enhancement proved that it is effective in malnutrition prediction even without anthropometric measurements, highlighting its potential to address malnutrition in constrained settings.