Talk:Random forest
From Wikipedia, the free encyclopedia
First time hearing about this. Reads a bit like an advertisement, and sort of short on info. I suspect that the claim of "most accurate classifier" would depend on the data being classified. I intend to read the link and see if I can contribute a bit. Dsol 09:34, 30 July 2005 (UTC)
- I agree, the choice of prediction algorithm depends on the nature of the data among many other factors, and claiming random forests are "most accurate" isn't really true. I've updated the article to make a less strong claim. Neilc 22:40, 2 April 2006 (UTC)
Here's one such specific application. Neslin, Gupta, Kamakura, Lu and Mason "Defection Detection: Measuring and Understanding the Predictive Accuracy of Customer Churn Models" Journal of Marketing Research (2006) Vol. XLIII p. 204-211. In a competition about predicting future customer termination, 44 entries compete for the best out-of-sample performance using a variety of methods. P.210: "Our winning entry used the power of combining several trees to improve prediction [...]". Innohead 11:24, 16 August 2006 (UTC)
