Session 4 - Advances in Probabilistic Modeling for Machine Learning: On Finite Mixture Modeling of Change-point Processes

Yana Melnykov, University of Alabama - Tuscaloosa

Presentation Type

Invited

Track

Other

Abstract

Change point estimation in standard process observed over time is an important problem in literature with applications in various fields. We study this problem in a heterogeneous population. A model-based clustering procedure relying on skewed matrix-variate mixture is proposed. It is capable of capturing the heterogeneity pattern and estimating change points from all data groups simultaneously. The appeal of such approach also lies in its flexibility to model the skewness and dependence in data with good interpretability. Two novel algorithms called matrix power mixture with abrupt change model and matrix power mixture with gradual change model are developed. The approaches are illustrated by simulation studies across a variety of settings. The models are then tested on the United States crime data with promising results.

Start Date

2-11-2020 9:30 AM

End Date

2-11-2020 10:30 AM

This document is currently not available here.

Share

COinS
 
Feb 11th, 9:30 AM Feb 11th, 10:30 AM

Session 4 - Advances in Probabilistic Modeling for Machine Learning: On Finite Mixture Modeling of Change-point Processes

Campanile & Hobo Day Gallery (A & B)

Change point estimation in standard process observed over time is an important problem in literature with applications in various fields. We study this problem in a heterogeneous population. A model-based clustering procedure relying on skewed matrix-variate mixture is proposed. It is capable of capturing the heterogeneity pattern and estimating change points from all data groups simultaneously. The appeal of such approach also lies in its flexibility to model the skewness and dependence in data with good interpretability. Two novel algorithms called matrix power mixture with abrupt change model and matrix power mixture with gradual change model are developed. The approaches are illustrated by simulation studies across a variety of settings. The models are then tested on the United States crime data with promising results.