Adaptive Structure Concept Factorization for Multiview Clustering.

Area of Science:

• Machine Learning
• Data Mining
• Artificial Intelligence

Background:

• Existing multiview clustering methods often compute graph matrices independently, neglecting crucial correlations between different data views.
• This limitation hinders effective data integration and can lead to suboptimal clustering performance.

Purpose of the Study:

• To develop a multiview clustering method that jointly optimizes graph matrices by exploiting inter-view correlations.
• To enhance data integration and improve clustering accuracy by considering relationships across multiple views.

Main Methods:

• A concept factorization-based multiview clustering approach is proposed.
• The method adaptively correlates affinity weights across all views.
• It is designed to be applicable to datasets containing negative values, unlike some nonnegative matrix factorization methods.

Main Results:

• The proposed method demonstrates superior performance compared to state-of-the-art approaches.
• Effectiveness is validated through experiments measuring accuracy, normalized mutual information, and purity.

Conclusions:

• Jointly optimizing graph matrices by leveraging inter-view correlations is effective for multiview clustering.
• The developed method offers a robust solution for data integration and clustering, particularly for datasets with negative values.