Skip to main content

Dept. Seminar - Qing Li

Sep 10, 2018 - 4:10 PM
to Sep 10, 2018 - 5:00 PM

Qing Li
Industrial & Manufacturing Systems Engr
Iowa State University

 

Change-Points Detection in the Recurrent-Event Context via Bayesian Inference

The driving risk during the initial period after licensure for novice teenage drivers is typically the highest but decreases rapidly right after. The change-point of driving risk is a critical parameter for evaluating teenage driving risk, which also varies substantially among drivers. This talk will present two recurrent-event change-point models for detecting the change-points by clusters of drivers with similar risk profiles: the Bayesian finite mixture model (BFMM) and the Dirichlet process mixture model (DPMM). We assume that the event counting process is a non-homogeneous Poisson process with piecewise-constant  intensity functions. Markov chain Monte Carlo algorithms were developed to sample from the posterior distributions. The simulation study suggests that the DPMM is not seriously affected by changes in the model assumptions. It outperforms the traditional BFMM in detecting the correct number of clusters, assigning subjects to the correct cluster, and computational efficiency.  The DPMM was applied to the Naturalist Teenage Driving Study, which continuously recorded the driving data of 42 novice teenage drivers for 18 months using advanced in-vehicle instrumentation.  Application to the Naturalistic Teenage Driving Study identified three distinct clusters. The results of this research provide more insight in teenagers' driving behaviour and will be critical to improve young drivers' safety education and parent management programs, as well as provide crucial reference for the Graduated Driver Licensing regulations to encourage safer driving. The proposed methods have practical applications in many fields such as transportation, medical studies, reliability of products, and human behaviour.

 


Refreshments at 3:45pm in Snedecor 2101.