Data and Decision-Making: Recommender Systems and Personalised Medicine
Free Public Lecture
Carrillo Gantner Theatre
Sidney Myer Asia Centre
T: 8344 9754
In this talk, Professor Howard Bondell will discuss two topics associated with the use of data for decision-making.
First he will address recommender systems – the systems used by online shopping websites to recommend additional products for a customer to purchase, or by TV streaming services to suggest new shows for a person to watch based on their past ratings. One issue faced by such systems is that much of the data is typically 'missing'; for example, most people would only rate a small number of TV shows. This leads to 'informative missingness', where the fact that data is missing is a critical piece of information in itself. Professor Bondell will show how this can be usefully incorporated into a recommender system to improve its predictions.
Second, Professor Bondell will discuss personalised medicine, which is premised on the ability to prescribe useful patient-specific treatments. To make this feasible, it is crucial to identify the variables that can inform the optimal treatment for any given individual. The standard techniques used for variable selection typically optimise prediction performance, but this is not necessarily the same as optimising treatment assignment. Professor Bondell will explain the difference between the two and show a technique that is able to optimise treatment assignment and thus be more useful for decision-making for personalised medicine.
Professor Howard Bondell, Professor of Statistics and Data Science
Professor Howard Bondell
Professor of Statistics and Data Science
University of Melbourne
Howard Bondell is Professor of Statistics and Data Science at the University of Melbourne, Australia. He is also Adjunct Professor at North Carolina State University, USA. Professor Bondell received his PhD in Statistics from Rutgers University in 2005. He began his academic career in 2005 in the Department of Statistics at North Carolina State University, and moved to the School of Mathematics and Statistics at the University of Melbourne in 2018. His research interests include variable and model selection, robust estimation, quantile regression, nonparametric smoothing and regression, regularization and Bayesian methods. Professor Bondell was elected Fellow of the American Statistical Association in 2017.