Research Project Title:
Machine Learning of Treatment Outcomes from Medical Records
abstract:The focus of this project is algorithmically predicting a patient's response to a treatment, leveraging the abundance of data in the healthcare industry. Our dataset is provided by the Blue Cross Blue Shield Association. This data does not result from controlled, randomized clinical trials and lacks a ground truth on the correct treatment decision for each patient. Therefore, this is an unsupervised learning problem. We will apply time series and causal inference techniques to model the progression of a disease when reasoning about the potential outcome of a treatment policy or intervention, working to control for sporadic availability of data, confounders that influence both the treatment and the outcome, and variation in medical practices.
"I am interested in this SuperUROP because it will provide me with the opportunity to work in multiple machine learning paradigms and to put the theory I have been studying in my classes to practical use. I hope to learn the subtleties of causal inference and time series modeling of medical data while simultaneously improving my own implementation and research abilities."