On Fuzzy Clustering of Categorical Data
碩士 === 中原大學 === 數學系 === 87 === Abstract Described here are four approaches to estimating the parameters of a mixture of multivariate Bernoulli distributions. The first approach is the MLE method proposed by Goodman [4]. The second approach is based on the well-known expectation...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | en_US |
| Published: |
1999
|
| Online Access: | http://ndltd.ncl.edu.tw/handle/63014190246217721055 |
| Summary: | 碩士 === 中原大學 === 數學系 === 87 === Abstract
Described here are four approaches to estimating the parameters of a mixture of multivariate Bernoulli distributions. The first approach is the MLE method proposed by Goodman [4]. The second approach is based on the well-known expectation maximization (EM) algorithm. The third one is the classification maximum likelihood (CML) algorithm which was discussed by Celeux and Govaert [12]. In this paper, we propose the fourth approach by using the so- called fuzzy class model and then create the fuzzy classification maximum likelihood approach for binary data. The accuracy, and robustness of these four types of algorithms for estimating the parameters of the multivariate Bernoulli mixtures are compared by using real empirical data and samples drawn from multivariate Bernoulli mixtures of two classes.
|
|---|