Unsupervised Decision Forest for Data Clustering and Density Estimation

TL;DR

This paper introduces an unsupervised decision forest algorithm that combines Random Forest and Gaussian Mixture Models to improve clustering and density estimation, outperforming existing methods in robustness and efficiency.

Contribution

It presents a novel dual assignment parameter and a robust affinity graph construction method that enhances unsupervised clustering performance over spectral clustering.

Findings

Outperforms six state-of-the-art clustering methods

Effective in capturing complex data structures

Robust across diverse datasets

Abstract

An algorithm to improve performance parameter for unsupervised decision forest clustering and density estimation is presented. Specifically, a dual assignment parameter is introduced as a density estimator by combining Random Forest and Gaussian Mixture Model. The Random Forest method has been specifically applied to construct a robust affinity graph that provides information on the underlying structure of data objects used in clustering. The proposed algorithm differs from the commonly used spectral clustering methods where the computed distance metric is used to find similarities between data points. Experiments were conducted using five datasets. A comparison with six other state-of-the-art methods shows that our model is superior to existing approaches. Efficiency of the proposed model is in capturing the underlying structure for a given set of data points. The proposed method is also robust, and can discriminate between the complex features of data points among different clusters.