This is the final team project for AC209: Data Science 1. In this project, I worked with other three students to build a classifier for predicting death of patients admitted to the emergency department in a hospital in Sweden.
Data were collected from patients admitted to a large (>60,000 yearly visits) Emergency Department, including patient's demographics, lab test results, severity index, and hospital occupancy at the time the patient was admitted. Data are perturbed, but general trends remain.
We are interested in building a classifier to predict patient outcome and investigate the importance of variables in our classifiers. We started with several classification models in the sklearn package, and narrowed down to logistic and random forest for further parameter tuning. Once we came up with interesting models, we continued with error analysis. Our models misclassified ~25%-30% of the patients, and we decided to examine the characteristics of those patients that were misclassified to understand where our classifier fails.
Please visit the website for detailed information: Prediction of death in ER