Abnormal Behavior Detection in Farming Stream Data

Juliet Chebet Moso (1, 2) , Stephane Cormier (1) , Hacene Fouchal (1) , Cyril de Runz (3) , John Wandeto (2)

1 CRESTIC - Centre de Recherche en Sciences et Technologies de l'Information et de la Communication - EA 3804
2 DeKUT - Dedan Kimathi University of Technology
3 BDTLN - Bases de données et traitement des langues naturelles

Farmers have been able to detect, quantify, and respond to spatial and temporal variation in crops thanks to a variety of technological advancements in recent years. Precision farming aims to provide precise targeting of agricultural inputs while reducing waste and negative consequences. Precision agriculture technologies are effective instruments for increasing farm sustainability and production. These technologies provide ways to create more with less resources. Nowadays, improving agricultural production efficiency and crop yields is impossible without the use of contemporary digital technologies and smart machinery. The introduction of high-accuracy GPS technology into farm machinery, such as combine harvesters, has been a significant component of precision farming. In this paper, we present a streaming-based methodology for detecting anomalies in combine harvester GPS recordings. The key hypothesis is that, “similar points in a feature space have similar anomaly scores”. We examine a data-driven strategy with the goal of applying unsupervised detection algorithms to find anomalies on the fly. Based on the results of the experiments, we can conclude that LSCP beats all other strategies when the number of base detectors is varied. The AUCPR performance of LSCP with two base detectors (HBOS and MCD) is 8.02% better than the second best technique MCD.

Mots clés: Anomaly detection; Data streams; Precision farming; unsupervised learning

Domaines: Informatique [cs]; Sciences et techniques de l'agriculture

Juliet Chebet Moso, Stephane Cormier, Hacene Fouchal, Cyril de Runz, John Wandeto. Abnormal Behavior Detection in Farming Stream Data. International Conference on Smart and Sustainable Agriculture (SSA), 2021, Virtual conference, France. pp.44-56, ⟨10.1007/978-3-030-88259-4_4⟩. ⟨hal-03445069⟩