Seminar - Semi-supervised strategy of finite mixture models
SMS PhD Proposal
Speaker: Ying Cui
Time:
Monday 3rd October 2022 at 03:00 PM -
04:00 PM
Location:
Cotton Club,
Cotton 350
Groups:
"Mathematics"
"Statistics and Operations Research"
Abstract
This proposal introduces a semi-supervised strategy of finite mixture models that cluster the subjects for ordered categorical response data, such as those from the Likert scale response. We use the proportional odds model as the basic model structure, and propose semi-supervised clustering models for analyzing a dataset that has partial labeled clusters but lots of unlabeled ones. Model fitting is performed using the EM algorithm that incorporates some information about the known cluster membership to carry out a probabilistic clustering of the unlabeled ones. The simulation study and analysis of some initial results are also provided.
The link to Zoom session is
https://vuw.zoom.us/my/ecspostgrad