A novel data science application approach for classification of nutritional composition, instrumental colour, texture and sensory analysis of bael fruit (Aegle marmelos (L) correa)

Abstract

Bael, a sub-tropical fruit consumed extensively in South East Asia was dried under four different types of drying techniques namely sun, hot-air, microwave and freeze drying to preserve it for consumption in off season. The macro nutrient composition, instrumental colour and texture profile were analysed for fresh bael pulp as well as for dried products. Sensory analysis was carried out for all the samples studied. X-ray diffractogram (XRD) was studied to substantiate sensory attributes like hand-feel, mouth feel and overall acceptability. Principal component analysis (PCA), hierarchical cluster analysis (HCA) and a machine learning application namely self-organising map (SOM) were applied to classify the data obtained from experiments. PCA and HCA are two linear method used in food science though very limited application was reported for non-linear SOM method. Satisfactory results were obtained from all these novel data science procedures of classification. Linear correlation of factors considered was aligned with the clusters formed from the experimental data to study the interdependency of the proximate composition, quantified textural and colour parameters and sensory attributes. Sensory attributes like flavour, hand-feel and mouth-feel were strogly correlated (>0.75) with texrural, colour and nutritional factors analysed through state-of-art techniques.