Proceedings of the 1st International Conference on Artificial Intelligence, Communication, IoT, Data Engineering and Security, IACIDS 2023, 23-25 November 2023, Lavasa, Pune, India

Research Article

Artificial Intelligence Based Anomaly Detection in Patient Health Monitoring Using Ensemble Learning Methods

Download42 downloads
  • @INPROCEEDINGS{10.4108/eai.23-11-2023.2343250,
        author={Ajitha  P},
        title={Artificial Intelligence Based Anomaly Detection in Patient Health Monitoring Using Ensemble Learning Methods},
        proceedings={Proceedings of the 1st International Conference on Artificial Intelligence, Communication, IoT, Data Engineering and Security, IACIDS 2023, 23-25 November 2023, Lavasa, Pune, India},
        publisher={EAI},
        proceedings_a={IACIDS},
        year={2024},
        month={3},
        keywords={artificial intelligence semi-supervised k-means health care isolation forest unsupervised anomaly detection},
        doi={10.4108/eai.23-11-2023.2343250}
    }
    
  • Ajitha P
    Year: 2024
    Artificial Intelligence Based Anomaly Detection in Patient Health Monitoring Using Ensemble Learning Methods
    IACIDS
    EAI
    DOI: 10.4108/eai.23-11-2023.2343250
Ajitha P1,*
  • 1: Associate Professor, Department of Software Systems & Computer Science(PG) KG College of Arts and Science,Coimbatore,Tamilnadu
*Contact email: ajitha.p@kgcas.com

Abstract

A Novel anomaly detection algorithm called SemiAI-AnomalyDetect+ is proposed in this paper. It is specifically designed for large-scale patient datasets in the healthcare domain. Combining unsupervised K-means clustering and semi-supervised learning techniques, the algorithm achieves robust and adaptable anomaly detection. Its performance is evaluated on a diverse patient records dataset from the MIMIC-III critical care database, featuring various health-related attributes. Comparative analysis with anomaly detection algorithms, including Isolation Forest and One-Class SVM, revealed that SemiAI-AnomalyDetect+ outperforms the baselines in precision, recall, F1-score, and ROC-AUC. With an average precision of 0.86 and an ROC-AUC of 0.93, the proposed algorithm excels at identifying anomalies with greater accuracy and efficiency. The integration of a feedback loop and active learning mechanism allows it to continually improve, effective tool for anomaly detection in healthcare, addressing the dynamic challenges of patient data.