A Computational Framework for Nutrient Density Assessment and Food Categorization

Abstract

In this paper, we aim to explore the potential of clustering in creating a nutritional map of different foods based on the nutritional elements present in the food. We evaluated two clustering algorithms: DBSCAN (Density-Based Spatial Clustering of Applications with Noise) and Agglomerative Clustering. We used these algorithms to cluster the different foods in the Kaggle Food-101 dataset, which contains nutritional features such as proteins, carbohydrates, fats, and energy density. In order to enhance the efficiency of the clustering process and reduce the complexity of the data, we used the PCA (Principal Component Analysis) technique for data reduction. The Agglomerative Clustering technique with PCA demonstrated superior clustering quality compared to DBSCAN. This was based on the fact that the Agglomerative Clustering technique with PCA produced a higher Silhouette Score (0.41) and Calinski-Harabasz Index (85.97) than the DBSCAN technique. In our research, it was also found that the clusters produced by the Agglomerative Clustering technique with PCA could separate the different foods based on nutritional elements, which included high-protein foods, high-carb foods, and balanced diet foods.