Open AccessAn Improved under Sampling Approaches for Concept Drift and Class Imbalance Data Streams using Improved Cuckoo Search Algorithm
Author(s) -
Tirupathi Rao Gullipalli
Publication year - 2021
Publication title -
türk bilgisayar ve matematik eğitimi dergisi
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.218
H-Index - 3
ISSN - 1309-4653
DOI - 10.17762/turcomat.v12i2.1945
Subject(s) - cuckoo search , concept drift , computer science , data stream mining , scalability , data stream , data mining , class (philosophy) , sampling (signal processing) , curse of dimensionality , field (mathematics) , cuckoo , machine learning , artificial intelligence , evolutionary algorithm , mathematics , particle swarm optimization , telecommunications , zoology , filter (signal processing) , database , biology , pure mathematics , computer vision
One of the biggest challenges in the recent times in the field of data stream learning is to mitigate the presence of concept drift. There are numerous challenges in overcoming the concept drift, such as changing class ratio, huge volume of data and real time processing for effective knowledge discovery. Evolutionary search techniques are one of the new paradigms to handle huge dimensionality and scalability of the data streams. One of the finest and least applied evolutionary search approaches is the cuckoo search technique for data streams. To solve both the concept drift and class imbalance issues simultaneously, in this paper we have proposed an approach using nature inspired evolutionary optimizing technique known as Cuckoo Feature and Instance Selection (CFIS) algorithm. The performance evaluation of the proposed approach is done on an exclusive experimental setup of 15 data streams formed and compared with two data stream approach. Moreover, a set of six evaluation criteria’s are considered for showing overall better performance of the proposed approach in the presence of concept drift and class imbalance.