• Title: “Data Mining Approach for Automatic Discovery of Livestock Behavior Concerning Animal Welfare” ; Inria.
  • PhD student: Lucie Lepetit.
  • Affiliation: Inria, LACODAM.
  • Co-supervision:
    • Alexandre Termier (Inria).
    • Peggy Cellier (INSA).
  • Doctoral school: École doctorale Mathématiques, télécommunications, informatique, signal, systèmes, électronique.
  • Project duration: 2023–2026.

Project summary:

Agriculture faces increasing societal expectations, particularly regarding animal welfare in livestock farming. Current animal welfare indicators are based on the provision of resources such as space and food.

Project objectives:

• Simultaneously analyze two types of sensor data to analyze animal welfare in real-time: physiological sensors close to the animal (temperature, accelerometers, hormonal balance) producing high-frequency time series, and activity reports from sequences of categorical events from video data.
• Identify both strongly and weakly expressed behaviors and the weak signals preceding these behaviors.

Methodologies:

• New approaches to extracting interesting patterns adapted to these types of data, particularly extending the "Exceptional Model Mining" method to time series data. The interest of these approaches lies in having both a flexible pattern extracted from time series and a qualification of this pattern from descriptive attributes.