Date/Time Date(s) - 30/01/20243:30 pm - 4:30 pm
Location: KTH B105 (different location from Fall 2023)
Date/Time: Tuesday, January 30, 2024, 3.30 – 4.30 p.m. (will bring refreshments to KTH B105)
Speaker: Eman Alamer (McMaster University)
Title: Mixture Models for Clustering Mixed-type Data
Abstract: A mixture model is a powerful tool for modelling data in the model-based clustering paradigm. There have been many well-established mixture models to analyze either categorical or continuous data in recent years; however, little work has been done on mixed-type data, especially when skewness or outliers are exhibited in mixed-type data. Two mixture models will be presented for clustering mixed-typed data. The first mixture is for analyzing asymmetric mixed-type data. Herein, we used skew-t distribution for modelling the continuous variables along with latent variable models. The second model helps to identify and detect potential atypical points (outliers) in mixed-type data. In this model, we combine latent variable models with the contaminated normal distribution. Real and simulated data are used for illustration.