Multinomial Cluster-Weighted Models for High-Dimensional Data

TL;DR

This paper introduces a new mixture model called Multinomial cluster-weighted model (MCWM) for high-dimensional data classification, demonstrating its effectiveness through simulations and real data with superior clustering performance.

Contribution

The paper develops the MCWM, establishes its identifiability, and proposes estimation algorithms, advancing clustering methods for high-dimensional categorical data.

Findings

MCWM achieves high accuracy in clustering tasks.

The model outperforms existing methods based on AUC and accuracy.

Effective model selection criteria are identified.

Abstract

Modeling of high-dimensional data is very important to categorize different classes. We develop a new mixture model called Multinomial cluster-weighted model (MCWM). We derive the identifiability of a general class of MCWM. We estimate the proposed model through Expectation-Maximization (EM) algorithm via an iteratively reweighted least squares (EM-IRLS) and Stochastic Gradient Descent (EM-SGD). Model selection is carried out using different information criteria. Various Adjusted Rand Indices are considered as a different measure of accuracy. The clustering performance of the proposed model is investigated using simulated and real datasets. MCWM shows excellent clustering results via performance measures such as Accuracy and Area under the ROC curve.