Surprisingly we haven’t yet received a single solution for the April-May 2015 Challenge. Nowadays there are so many papers and posts about predictive analytics and… no solutions for a relatively simple problem? If not vendors then consultants could at least use some open source machine learning/data mining tools but they did not so far. Maybe the problem is more complex than we initially expected or the provided data is not sufficient… Could anybody share their thoughts?
Just in case, here is a link for a new article “Top 10 data mining algorithms in plain English“. Some popular data mining tools can be found in this Catalog.
Finally today we received the first solution submitted by Riccardo Hertel who used R and a naïve Bayesian model. His solution states:
“Patient 21’s illness is likely to be a cold – with a probability of somewhat above 83%, or an allergy, but the latter only with a probability of 12.4%. The probability of a flu is estimated to be very small in this case, with merely 3.8% “