Applying the Long-Term Memory Algorithm to Forecast Loss of Thermoregulation Capacity in Honeybee Colonies

• Antonio Braga UFC
• Lia Furtado UFC
• Antonio Bezerra UFC
• Breno Freitas UFC
• Joseph Cazier ASU
• Danielo Gomes UFC

DOI: https://doi.org/10.5753/wcama.2019.6422

Abstract

Bees are the main pollinators of most wild and cultivated plant species, thus being essential for the maintenance of plant ecosystems and for food production. But they are threatened due to a series of drivers such as pesticides, habitat loss and climate change. Here, we propose a method to iden- tify the loss of thermoregulation capacity in honeybee colonies. We applied the Long Short-Term Memory (LSTM) algorithm, which is based on Recurrent Neural Networks (RNN), to six real datasets of the Arnia remote hive monitoring system. From brood temperatures gathered along the European fall season in 2017, the LSTM was able to detect when a honeybee colony is about to lose its thermoregulation capacity. Our results showed an error of only 0.5% in predic- tion for well-thermoregulated beehives.

Keywords: Apis mellifera, monitoring, homeostasis, LSTM, forecasting