Keywords: flavonoids, chocolate, antioxidant, multivariate statistics, clustering
People like eating chocolate, but have concern on health risk, especially with Cardiovascular or Neurovascular disease. Multivariate Statistics was applied to define a health index on choosing a healthy chocolate for heart disease patients. Chocolate, made from cocoa beans, contains flavonoids. Flavonoids are the most abundant polyphenols in human diet. Polyphenols have antioxidant properties which can prevent aging and is beneficial to prevent heart disease and diabetes. Cocoa flavanols promote healthy blood flow circulation from head to toe. The heart, brain, and muscle depend on a healthy circulatory system. Data has been collected on 20+ chocolate ingredient contents from 60+ different types of chocolate. Multivariate correlation has found that (1) strong negative correlation between Cocoa and Sugar, and (2) strong positive correlation between Diet Fiber and Iron. Dark chocolate contains more cocoa, and less sugar. Dietary fiber and iron are strong in correlation because of the high cocoa. Hierarchical Clustering can divide chocolates to three clusters: Dark, Milk, and White. The Cocoa and Calcium are the most deciding factors to separate three clusters. Based on Chocolate Science, Heart Disease Risk, and Cluster Analysis, a Health Index has been defined as [(Cocoa% * 2) + (Dietary_Fiber * 2) - (Sugar * 2) + Calcium – (Saturated_Fat * 2) – (Cholesterol * 2) – Sodium*1 + Iron*1]. The weighting coefficients are assigned based on heart disease risk. People with heart diseases should eat less of saturated fat, trans fat, sodium, and cholesterol. Each variable raw data has been transformed to Z-standardized scale to remove any larger variance bias. Among 60+ chocolate products analyzed, the products with highest index all have >80% cocoa, 0mg cholesterol, higher dietary fiber, and lower sugar. These healthiest chocolates may prevent heart disease, cancer, and diabetes. The Multivariate Statistics can interpret Sciences in a predictive/empirical modeling.