Classifying Apis mellifera Breeds Using Data Mining Techniques Based on Morphological Traits

Abstract

The classification of bee breeds is significant for breeding, maintaining genetic diversity, increasing productivity and protecting the health of the bee colonies. Therefore, this study aims to classify different honeybee breeds based on their morphological traits using data mining techniques, which are cost-effective and straightforward. It were used a total of 35 colonies from a private bee farm for morphometric analysis in the study, which included seven different bee breeds and 404 bee samples. A range of data mining techniques (Support Vector Machines (SVM), Random Forest (RF), Artificial Neural Networks (ANN), Naive Bayes (NB) and k-Nearest Neighbors (k-NN)), and model fit criteria were used for the classification of bee breeds. Overall, the study shows significant differences in the morphological traits of different bee breeds, highlighting the diversity and different traits of each bee breed. In addition, the study shows that the RF model is superior in all criteria and therefore the most effective for classifying honeybee breeds. In contrast, the NB model consistently performs the worst, as evidenced by the consistently minimum values of all metrics. In conclusion, RF model exhibiting a 99.8% success rate, stands out as highly effective in the classification of bee breeds based on the morphological traits, supporting its applicability in future classification research.