Talk:Predictive analytics/Archives/2014

Predictive analytics vs. predictive modelling

Predictive analtytics, a sister science to predictive modelling, is seperate and distinct as it is an insurance industry term which specifically refers to making realtime live judgements about the likelihood that a particular insured is going to have an accident or is lying on an insurance claim. —Preceding unsigned comment added by 69.27.2.186 (talk • contribs)

Yes, I left a comment in the redirect to that effect. I've put the above text as a stub for now instead of the redirect, although I think the field of predictive analytics is broader than that insurance industry view. – Doug Bell ^{talk•contrib} 20:33, 21 February 2006 (UTC)

It really looks to me like this predictive analytics entry and predictive modeling are largely covering the same topics. And it's odd that the predictive modeling entry doesn't contain any links to predictive analytics. To me, it looks like these two wikipedia entries should be combined. My idea would be to incorporate some of the material from the predictive modeling entry into the predictive analytics entry, then establish a "re-direct", so that when someone types in "predictive modeling" they are brought to predictive analytics. However, I'm an inexperienced wikipedia contributor, and am unaware of the general wikipedia conventions for consolidating articles like this. I also do not want to offend the authors who have put lots of hard work into the two entries. -- So I merely want to bring up the idea here for discussion to see what others think of the idea. Karl (talk) 15:34, 1 March 2014 (UTC)

I agree. When I edited the predictive modeling I did not realize this page existed. Seeing it, I realize that the amount of overlap is so substantial that the two need merging. I'm also too much of a novice to know where to start. We could expand the section on predictive models in this page to include the models from the predictive modeling page. Currently on this page they are under "Analytical Techniques" which I think is vague and generic. Hence a Models section would make more sense. — Preceding unsigned comment added by Thomas Speidel (talk • contribs) 23:20, 1 March 2014 (UTC)

Criticism

The criticism section lacks context in that the example reported deals with the difficulties of predicting human behaviours. Yet, a predictive model can be used for any quantifiable measure, not just behaviours. I think criticism also needs to include the problems that make a model inappropriate. For example, statistical literacy in applying the right methodology, understanding the limitation of each methodology, having realistic assumptions, diagnostics, the treatment of missing information, measurement error, several types of bias, internal and external validation, overfitting.

Tools

The tools section appears biased and somewhat missinformed. Thus I have tagged two statements with citation needed. In particular" "However, modern predictive analytics tools are no longer restricted to IT specialists". It is not clear how predictive analytics tools used to be a prerogative of IT specialists. The other statement: "Predictive analytics tools have become sophisticated enough to adequately present and dissect data problems" also implies that these tools were not sophisticated enough in the past, which I do not believe is accurate. I think the confusion might be caused by the popularity some of those tools are now enjoying which makes them look "new".