Search papers, labs, and topics across Lattice.
This paper explores the novel application of meta-classification to one-class classification (OCC) models by representing them as normality rankings and employing nearest-neighbor and ranking-correlation metrics for classification. The approach successfully classifies OCC models based on training datasets, algorithms, and hyperparameters, achieving high accuracy particularly when class labels correspond to datasets. The findings suggest that classifying OCC models can effectively be viewed as classifying datasets, offering a unified framework for understanding and organizing machine learning models.
Classifying one-class classification models reveals that dataset characteristics can significantly influence model performance, challenging traditional views on model evaluation.
Machine Learning (ML) techniques have been applied to various problems. However, applying ML to ML models is an unexplored direction. For this purpose, this paper considers a meta-classification of one-class classification (OCC) models, because all ML models could be approximated as OCC models. The proposal represents OCC models as normality rankings and classifies them using nearest-neighbor and ranking-correlation metrics. The experiment classifies OCC models, where classes correspond to training datasets, algorithms, and hyperparameters. The proposal achieves high accuracy when class labels are datasets. Moreover, it can classify algorithms when the training datasets contain the same class. In addition, the discussion highlights that the classification of OCC models is essentially the classification of datasets that treats multiple samples as a single input. The experiment demonstrates the classification of datasets using sleeping records. The proposed method can provide a unified solution for classifying OCC models, datasets, and rankings. Source code is uploaded to the public repository https://github.com/ToshiHayashi/ClassOCC.
