Classification of edible bird’s nest samples using a logistic regression model through the mineral ratio approach

Abstract

Cave-harvested edible bird nest (EBN) is a high-priced commodity, that often being counterfeited with lower-priced house-farmed EBN. In this study, cave-harvested EBN and house-farmed EBN were classified based on to the concentration of calcium (Ca), sodium (Na), magnesium (Mg) and potassium (K) present. To solve the convergence failure problem caused by the complete separation of the EBN data, a logistic regression model analysis on 48 EBN samples harvested from Malaysia and Indonesia through a mineral ratio approach was adopted. Out of the 3 logistic regression models developed, the model consisting of Ca/Na ratio and Mg/K ratio gave the best performance showing no convergence failure of Maximum Likelihood Estimation (MLE) and both the explanatory variables were highly significant. The result indicated that both Ca/Na ratio and Mg/K ratio, affecting the probability of EBN type to be cave-harvested EBN in a positive manner. The logistic regression model developed with the Ca/Na ratio and Mg/K ratio gave a 100% specificity and 91.67% sensitivity in classifying the EBN type. The results of the analysis were verified using the Receiver Operating Characteristics curves. The validation result indicated that the model has a very good overall diagnostic accuracy in classifying the EBN type based on the mineral ratio.