Special thanks for mentioning Credit Score. The algorithm we are building is pretty much like a Credit Score algo. Not just because we show a score for users, but because we use factors learned and deduced from similar other profiles and use it to quantitatively predict the likelihood of a fit (risk in case of a credit score). It would be ideal to use larger datasets for training, however its still in beta and the algorithm is being updated continuously.
Thanks for the clickable link too :)