An optimized sliding window approach for Concept Drift Detection and adaption

Due to dynamic smart systems, concept drift in live streaming data is a typical issue, resulting inperformance reduction. Despite the fact that there are a variety of traditional ways for handling streaming data, they have all failed to address notion drift, necessitating the development of an adapt...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:NeuroQuantology 2022-01, Vol.20 (11), p.3872
Hauptverfasser: Ketan Sanjay Desale, Shinde, Swati
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Due to dynamic smart systems, concept drift in live streaming data is a typical issue, resulting inperformance reduction. Despite the fact that there are a variety of traditional ways for handling streaming data, they have all failed to address notion drift, necessitating the development of an adaptive approach to manage dynamic streaming data. As a result, a new strategy for dealing with idea drift difficulties in online data streaming is presented in this study. This study creates a dynamic streaming data analysis system based on an optimized Deep CNN and an optimized adaptive and sliding window (OASW) technique that efficiently tackles memory and time restrictions.For offline learning, an optimized Deep CNN classifier is used as a base classifier, which is created by combining the proposed aggressive hunt optimization (AHO) method with the Deep CNN classifier to tune the classifier's ideal parameters. In this study, an optimized adaptable and sliding window is used to adjust pattern changes in data streams, successfully handling concept drift. The proposed methods exceed the traditional methods in terms of specificity, sensitivity, accuracy, F1 score, and precision score of 94.55%, 95.78%, 95.98%, 96.47% and 96.87%, respectively, according to the experimental study.
ISSN:1303-5150
DOI:10.14704/nq.2022.20.11.NQ66390