Need a Classification Algorithm or 17?
March 21, 2016
I gave a lecture a couple of years ago about the similarity among major content processing systems. In that talk, I focused on 10 numerical recipes which our research identified in the commercial products from a number of well known intelligence platform vendors. The point of the lecture was to underscore the baked in weaknesses of platforms which use procedures taught in many universities. Outputs often vary because of the goofy decisions humans make or because the underlying data pumped into the numerical recipes is flawed.
I want to call your attention to “Implementation of 17 Classification Algorithms in R.” If you want to see the differences classification algorithms output, just fire up your system, implement these 17 methods, and check out the results. Our research reiterated to my goslings that one can select a classification algorithm to produce the type of output desired by the system engineer. Yep, put your hands on the steering wheel and drive that output pretty much where you want it to go. Do users of content processing systems know about these baked in pre-loaded destinations? Nah.
Stephen E Arnold, March 21, 2016
Comments
-
- Subscribe to Beyond Search
Feature archive
News archive
-
