Senger, DirenKluss, ThorstenFörster, AnnaWohlgemuth, VolkerKranzlmüller, DieterHöb, Maximilian2023-12-152023-12-152023978-3-88579-736-4https://dl.gi.de/handle/20.500.12116/43329Beekeepers in most parts of the world are challenged by colony losses induced by diseases, parasites, shortage of nectar and pollen, and various other causes. For a better understanding of these causes and to inform beekeepers when to intervene and to perform certain beekeeping activities to protect their colonies, monitoring systems using sensor technology in the hives can be implemented. Currently, most monitoring systems available at the market provide a visualisation of the measured sensor values, but do no integrate further analysis or an interpretation of the values, e.g. by time series classification or by comparing to time series prediction data. We describe a system architecture where predictions made for a specific colony can be used to find aberrations, potentially indicating an anomalous development of the bee colony. We summarise challenges of such an implementation and evaluate the system using data from a German Citizen Science Project, consisting of temperature, humidity and weight measurements and a log of all activities and observations made by the beekeepers in a web app.ensensor nodespredictive modelsIoTagricultureenvironmental modellingText/Conference Paper10.18420/env2023-0011617-5468