Online Data Stream Clustering: Proposal
Résumé
Pattern recognition methods, specially classification, has been growing in popularity because its ability to adapt in a changing environment. This paper proposes a classification of such techniques found on a literature review of dynamic classification techniques collected among a variety of fields, including fault detection, identification of moving objects, bank customer segmentation, among others. Based on how the dynamic behavior is incorporated in the classification process (data, classifier structure or both), three main categories are detected: Methods for classifying static objects using dynamic classifiers, methods classifying dynamic objects with static classifiers and finally methods classifying dynamic objects with dynamic classifiers.
Origine | Fichiers produits par l'(les) auteur(s) |
---|
Loading...