Model Based Clustering using Finite Mixture Models of Lognormal Distribution

Abstract

Abstract
Model-based clustering consists of fitting a mixture model to data and identifying cluster with its components. This study discusses model-based clustering using a finite mixture of a univariate lognormal distribution with k components. In this study, lognormal mixture models for skewed data and also the estimation of mixing proportion by distance-based clustering have been considered. Parameter estimation is carried out by expectation maximization (EM) algorithm using different initialization approach using Manhattan distance, k-means and random methods. This clustering approach has been illustrated with the help of both simulated and real clustering datasets.
Keywords: Finite mixture model, model-based clustering, Lognormal Distribution, EM algorithm, mixing proportion, K-means algorithm

Cite this Article

Deepana R., Kiruthika. Model Based Clustering using Finite Mixture Models of Lognormal Distribution. Research & Reviews: Journal of Statistics. 2018; 7(3): 58–67p.