JCSE, vol. 15, no. 1, pp.1-17, 2021

DOI: http://dx.doi.org/10.5626/JCSE.2021.15.1.1

A Knowledge Extraction Pipeline between Supervised and Unsupervised Machine Learning Using Gaussian Mixture Models for Anomaly Detection

Reda Chefira and Said Rakrak
Laboratory of Engineering Informatics and Systems at Faculty of Sciences and Techniques, Cadi Ayyad University, Marrakech, Morocco

Abstract: This paper presents a new approach to design a decision support model with suitability across various contexts, and in particular for the Internet of Things. It provides an anomaly detection-learning model that is adapted to the patient's medical condition. A highly balanced artificial intelligence based on a Gaussian mixture model and association rules leverages the knowledge acquired through cross-referencing supervised and unsupervised machine learning. This process ensures an unsupervised cluster-based model, to accurately classify medical inputs according to their risk level and provide a knowledge extraction bridge between the supervised and unsupervised aspects of the data, thereby enhancing the medical decision-making process to be data-driven and therefore case-specific.

Keyword: Classification; Clustering; Association rules; Knowledge extraction; Swarm intelligence

Full Paper:   255 Downloads, 1509 View