Predicting Time Spent with Physician

The goal of this project is to predict how much time a physician should be spending with a patient, given the patient’s injury, referral status, and other reported, nonidentifiable medical information. This may be used as an additional factor for measuring and predicting physician efficiency alongside existing methodology. We implement linear regression and a variety of classification techniques on a national dataset of Emergency Room (ER) patients and outpatients. We use cross validation to select the optimum set of features. In addition, we implement the k-Means algorithm to generate new features for our supervised learning algorithms. Our final models provide good predictions of time spent with physician with a mean error of around 7 minutes.