Anomaly Detection on Univariate Sensing Time Series Data for Smart Aquaculture Using K-Means, Isolation Forest, and Local Outlier Factor

Aquaculture has a great importance in economic development and food production. Maintaining an ecological environment with good water quality is the most critical link to ensure the efficient and qualitative of aquaculture. Good management of the water quality can avoid occurrence of abnormal conditions and significantly contribute to secure food in the future. Detection of anomalies ensures that the aquaculture environment is maintained properly to meet healthy and proper requirements for fish farming. The main focus of this paper is the use of machine learning approaches to detect anomalies for water quality data in aquaculture environment. It presents an analysis of three machine learning anomaly detection techniques: the K-Means clustering, the Local Outlier Factor, and the Isolation Forest. Extensive analysis of the mentioned techniques was conducted using several sensor datasets obtained from a real-world IoT aquaculture system, specifically for the parameters of temperature, dissolved oxygen, and pH. The evaluation analysis reveals that K-Means and Isolation Forest anomaly detection methods show promising results in detecting anomalies for the three aquaculture parameters.