Time Series Clustering for Smart Beehives
26.11.2024Smart beehives collect sensor data that reflects developments of and events in the bee colony, such as hive growth, successful nectar collection and beekeeper interventions. None come with labels. Therefore, we use different clustering algorithms on the time series data to find patterns in the data.
Since 2019, we have collected data from smart beehives in two projects: we4bee and BeeConnected. We4Bee aimed at developing algorithms to gain new insights into bee colonies, BeeConnceted researched winter mortality of honey bee colonies. In both projects, we have collected weight data among other sensor data such as temperature. These data time series contain patterns that allow ecologists insights into the development of the hive.
As all the data comes without labels, we use clustering algorithms to obtain patterns that describe the behavior of the hive. In ongoing research, data has been clustered using standard clustering and time serires clustering algorithms.
Your tasks:
• train clusterings of univariate and multivariate data
• speed up clustering
• deploy other methods to identify motifs and discords in the time series
Your profile:
• student of computer science, maths, physics, natural sciences
• ideally experience in programming python
• benficial: experience with git
• beneficial: experience with scikit-learn and aeon
