Drift detection and model selection algorithms: concept and experimental evaluation

Authors:
Piotr Cal;Michał Woźniak
Affiliations:
Department of Systems and Computer Networks, Wrocław University of Technology, Wrocław, Poland;Department of Systems and Computer Networks, Wrocław University of Technology, Wrocław, Poland
Venue:
HAIS'12 Proceedings of the 7th international conference on Hybrid Artificial Intelligent Systems - Volume Part II
Year:
2012

Citing 7
Cited 0

A streaming ensemble algorithm (SEA) for large-scale classification

Proceedings of the seventh ACM SIGKDD international conference on Knowledge discovery and data mining
Machine Learning

Machine Learning
Combining Pattern Classifiers: Methods and Algorithms

Combining Pattern Classifiers: Methods and Algorithms
Pattern Recognition and Machine Learning (Information Science and Statistics)

Pattern Recognition and Machine Learning (Information Science and Statistics)
Adaptive concept drift detection

Statistical Analysis and Data Mining - Best of SDM'09
Introduction to Machine Learning

Introduction to Machine Learning
Artificial recurrence for classification of streaming data with concept shift

ICAIS'11 Proceedings of the Second international conference on Adaptive and intelligent systems

Quantified Score

Hi-index	0.00

Visualization

Abstract

The traditional classifier cannot keep its quality, when the concept drift appears. The paper proposes how to protect against classification quality decreasing when concept drift occurs. Invented methods do not train classifiers all the time but they try to use earlier gained knowledge about models and switched older model to suitable new one. In this work we assume that the set of models is known and stored as the pool of classifiers. Then, by using drift detecting and searching models methods, we can choose the best model. Our propositions and the main characteristics of them were evaluated on the basis of the experiments which were carried out on chosen artificial data set.