Machine learning platform for beehive health monitoring

When honeybees (Apis mellifera L.) are compared with other insects, they are known for helping the environment and other animals. The significance of bees goes beyond just producing honey that are consumed by humans for thousands of years. They also produce beeswax, which is used in a variety of products, including candles, cosmetics, and food. Other than production, they are important pollinators for a variety of crops, including fruits, vegetables, and nuts. By pollinating, they help to increase crop yields, and improve the quality of produce. Recently, various stress factors have contributed to a decrease in honey bee colonies such as climate change, chemical usage in farming, parasites, diseases and loss of meadows and forests. With monitoring, early detection of beehive status could be done and early detection is important for the survival of the beehive. In this context, analyzing the sound produced within bee hives is a crucial method for non-intrusive monitoring. This paper focuses on queen detection and swarm activity detection by mainly looking the sounds that are produced. For this purpose, two different audio extractor techniques are used, MFCC(mel frequency cepstral coefficients) and STFT (short time fourier transform). After the extraction operation, Neural Network and SVM algorithms are used for the classification. Most of the results are higher than 90\%. The best case for Neural Network with MFCC extractor is 96\% accuracy and best case for SVM with MFCC extractor is 97\%.