The Heritage Health Prize is a contest for predicting how long people will stay in the hospitals based on medical data. The top prize is for $3 million dollars but there are large milestone prizes that are awarded after a certain period of time.
Details for the Prize can be found here:
To participate in the contest, you will need to write computer code but the good news is that there are two pieces of information that could make this prize worth pursuing even if you do not have any medical or data mining background.
First, as far as I understand it, you are allowed to use open source code. If you go back to the Netflix Prize forum http://www.netflixprize.com//community/ , for example, you can find links to lots of data mining code.
Second, each milestone winner will publish their results in clear enough language that their algorithm and formula can be used by everyone else. This is also what happened with the Netflix Prize where one of the teams that won had no data mining experience previously. The team known as Pragmatic Theory did just that. They learned all their insights from the published milestone prizes (see http://en.wikipedia.org/wiki/Netflix_Prize ).
If the same methods work on this prize as worked on the Netflix Prize, then the goal will be to learn the following algorithms:
(1) Global Effects (see http://algorithmsanalyzed.blogspot.com/2008/05/bel... )
(2) Singular Value Decompostion (see http://en.wikipedia.org/wiki/Singular_Value_Decomp... )
(3) Restricted Boltzmann Machine (see http://en.wikipedia.org/wiki/Restricted_Boltzmann_... )
All the details on the winning solution for the Netflix Prize are here: http://www.netflixprize.com//community/viewtopic.p...
That's pretty much it. Good luck. Oh, and if you are one the winners, please remember this answer on HubPages. :-)