Identifying bee species origins of Philippine honey using X-ray fluorescence elemental analysis coupled with machine learning

Abstract

Stingless bee honey is emerging as a superfood, given its enhanced health and therapeutic benefits. In this paper, we used handheld X-ray fluorescence spectroscopy (hXRF) with machine learning techniques to classify Philippine honey based on its entomological origin. Honey samples from three different bee species were analyzed, specifically European honeybee (Apis mellifera), Philippine giant honeybees (Apis breviligula and Apis dorsata), and Philippine stingless bee (Tetragonula biroi). Random forest and logistic regression models were used on the hXRF dataset for entomological origin classification. The optimized random forest model classified entomological origin with 85.2 % (225 out of 264) overall accuracy. The logistic regression model confirmed the entomological origin of Philippine stingless bees with 94.1 % accuracy and 100.0 % specificity. As such, honey that passes this model's test is undoubtedly made by Philippine stingless bees, making it an excellent screening tool for authenticating Philippine stingless bee honey.