Is kaggle planning to submit some benchmark entries, as has been done in other competitions?  Elsewhere in the forums is discussion of a zero benchmark, a 15 benchmark, and a constant-value benchmark.

Personally, I think Jeremy Howard's best shot at this competition would be an excellent benchmark, and since he works for kaggle he could eventually release his code for all of our benefit.

I'm planning to release some tutorials over the coming months on developing models using the HPN data. I will submit the models I come up with when developing these tutorials. So that will give you a good feeling for what is required to get particular results. I'm hoping to have the first of these available a few weeks after the remaining data is released.

Awesome!

I would like to know whats the current RMSLE score HPN has on its own algorithms. i.e. like Netflix published their score, that is if HPN already has some analytics running on this data.

We can't give you an HPN benchmark because they've not tackled this problem before. 

Out of curiosity then how did the benchmark become .4 , why pull .4 out of the box?

Because they want to spend 500k rather than 3m?

My Thanks was more of a Facebook Like as it made me laugh!

 

This would be very much appreciated. I'm now watching An Introduction to Medicare Claims Public Use Files and it would be nice to have a couple of (hypothetical) examples that would show how entries for HHP_release were created.

Thank you very much!