Predictive analysis of health care data

OData support
Supervisor:
Gáspár Csaba
Department of Telecommunications and Media Informatics

For my Master's Thesis, I participated in the Heritage Health Prize, the most renown data mining competition in recent years. The goal of this competition was to predict future hospitalizations of individual patients, thus helping to provide better health care at lower costs.

Based on the competition's event and member information datasets, I calculated a large number of new powerful explanatory variables. This dataset was used for building several different types of predictive models, including blended models which combined predictions from several individual models.

Creation of new variables and model building were performed iteratively. By the end of the competition I had created 145 explanatory variables. The predictions of my best performing blended model secured me a final position of 65th out of 1659 teams.

Downloads

Please sign in to download the files of this thesis.