The Heritage Health Prize is a competition to find the best algorithm for the identification of patients which will be admitted to hospital in the next year using historical claims data.
The competition is organized by the Heritage Health Provider Network, a Californian Hospital Network. A prize money of 3 million dollar is being offered. Kaggle, a platform for data mining competitions, provides the infrastructure for the Heritage Health Prize. Over a period of two years data scientists did have time to train their solutions with the provided set of anonymized patient claim data. All solutions have been evaluated with an independent test set. Over 1,600 teams from all over the world participated in the contest.
As a member of team Combined Power, eonum has contributed an important part of the solution for the 5th place in the final ranking. The competition ended on 4 April 2013. The winner and the Top 10 ranks have been revealed on 3 June at the Health Datapalooza IV in Washington DC.
The techniques used for the competition included Long Short Term Memory (LSTM) Neural Networks, Gradient Boosting Machines and several Ensembling methods.
Heritage Health Prize: http://www.heritagehealthprize.com
Articles and News about the Heritage Health Prize: http://www.heritagehealthprize.com/news
Kaggle: Go from big Data to Big Analytics: http://www.kaggle.com/
Heritage Health Provider Network: http://www.heritageprovidernetwork.com/
Health Datapalooza: http://healthdatapalooza.org/