A Diversity Framework for Dealing With Multiple Types of Concept Drift Based on Clustering in the Model Space

Area of Science:

• Machine Learning
• Data Mining
• Artificial Intelligence

Background:

• Data stream applications frequently encounter diverse concept drift types.
• Existing methods often struggle with multiple drift types, limiting predictive accuracy.

Purpose of the Study:

• To propose a novel diversity framework for effectively managing multiple concept drift types in data streams.
• To enhance predictive performance by leveraging ensemble diversity for concept adaptation.

Main Methods:

• Utilizing a diversity framework based on clustering in model space.
• Building diverse ensembles to represent different concepts and accelerate adaptation.
• Identifying recurring concepts through model space clustering.

Main Results:

• The diversity framework demonstrates competitive or superior prequential accuracy compared to existing methods.
• Effectiveness is particularly pronounced with recurring concepts or when new concepts share similarities with past ones.
• Experiments conducted on 20 synthetic and 3 real-world data streams.

Conclusions:

• The proposed diversity framework offers a robust solution for handling multiple concept drifts.
• Diversity in ensembles aids in adapting to various drift types and managing recurring concepts efficiently.