Fast-RCM: Fast Tree-Based Unsupervised Rare-Class Mining.

Area of Science:

• Data Mining
• Machine Learning
• Pattern Recognition

Background:

• Imbalanced datasets pose challenges in identifying rare classes.
• Existing rare-class mining (RCM) methods often require significant labeled data, which is costly and time-consuming to obtain.
• Unsupervised RCM remains largely unexplored.

Purpose of the Study:

• To address the limitations of supervised RCM by investigating the unsupervised RCM problem.
• To propose an efficient algorithm for unsupervised RCM.
• To demonstrate the effectiveness and efficiency of the proposed algorithm on various datasets.

Main Methods:

• Developed Fast-RCM, an algorithm with approximately linear time complexity concerning data size and dimensionality.
• The algorithm constructs a 'rare tree' for the unlabeled dataset.
• Rare class instances are extracted based on the generated rare tree.

Main Results:

• Fast-RCM achieves approximately linear time complexity, significantly outperforming existing methods with quadratic or cubic complexity.
• Experimental evaluations on synthetic and real-world datasets confirm effective and efficient rare class extraction in unsupervised settings.
• The proposed algorithm demonstrates approximately five times the speed of state-of-the-art methods.

Conclusions:

• Fast-RCM offers an efficient and effective solution for unsupervised rare-class mining.
• The algorithm's scalability makes it suitable for large-scale datasets.
• This work pioneers unsupervised RCM, providing a valuable tool for analyzing imbalanced data without prior labeling.