Review on Machine Learning Models for Efficient Queen Bee Identification in Beehives

This paper proposes an acoustic monitoring technique for beehive health assessment by analysing sound spectra. Researchers created a database of audio recordings and developed a system for continuous acoustic surveillance. Two colony states—healthy (with a queen bee and normal activity) and queen less were simulated. Using the Mel-frequency cepstral coefficients (MFCC) method, distinctive frequencies were extracted from the spectral content of these phases. The system achieved hive health identification within 30 minutes by comparing specific acoustic traits. Traditional queen detection is time-consuming and disruptive, prompting this investigation into auditory data and machine learning (ML) for hive state classification, including queen absence. Support vector machines (SVM) and convolutional neural networks (CNN) were evaluated on data from the NU Hive Project, demonstrating promising results despite challenges in adapting to new hives. Omnidirectional microphones and Raspberry Pi-based systems were employed for monitoring. MFCC feature extraction and Lasso logistic regression were used for feature selection and regularization in a multiclass model. This work highlights the potential of ML in addressing queen less colony detection through sound pattern analysis.