Semi-Supervised Cascaded Clustering for Classification of Noisy Label Data

TL;DR

This paper introduces a semi-supervised cascaded clustering algorithm that effectively handles noisy labels in limited datasets, improving classification accuracy without extensive human intervention, especially suited for industrial applications.

Contribution

The paper presents a novel SSCC algorithm with a cluster evaluation matrix for noise reduction, reducing reliance on deep neural networks and human expertise in noisy, limited datasets.

Findings

Outperforms SVM on noisy datasets

Effectively identifies and eliminates noisy labels

Produces accurate classifiers with minimal human input

Abstract

The performance of supervised classification techniques often deteriorates when the data has noisy labels. Even the semi-supervised classification approaches have largely focused only on the problem of handling missing labels. Most of the approaches addressing the noisy label data rely on deep neural networks (DNN) that require huge datasets for classification tasks. This poses a serious challenge especially in process and manufacturing industries, where the data is limited and labels are noisy. We propose a semi-supervised cascaded clustering (SSCC) algorithm to extract patterns and generate a cascaded tree of classes in such datasets. A novel cluster evaluation matrix (CEM) with configurable hyperparameters is introduced to localize and eliminate the noisy labels and invoke a pruning criterion on cascaded clustering. The algorithm reduces the dependency on expensive human expertise for assessing the accuracy of labels. A classifier generated based on SSCC is found to be accurate and consistent even when trained on noisy label datasets. It performed better in comparison with the support vector machines (SVM) when tested on multiple noisy-label datasets, including an industrial dataset. The proposed approach can be effectively used for deriving actionable insights in industrial settings with minimal human expertise.