Humans: An essential ingredient for any machine learning project
Jul 27, 2020 6:48 AM
| Last Modified: Jul 29, 2020 11:17 AM
Julia O'Neill, Direxa Consulting, talks about the human element of machine learning projects in a recent panel discussion.Julia O’Neill, Direxa Consulting, explains why subject matter expertise is crucial for any machine learning project. You can see the full panel discussion in this on-demand video.
“Where humans are really essential in setting up data for machine learning or any sort of analysis are, first of all, deciding what components of the data will be useful for trying to formulate a model. And that sounds easy, it sounds like we should be able to just look down the list and pick out all the ones that are in and the ones that are out, but it gets really tough.
“First of all, we typically for these problems have many hundreds of potential predictor variables, and it’s just really hard to sift through that. And the rules are a little different with some of these multivariate methods. In a simple linear, univariate or just-a-few-variables-at-a-time world, we would pick out the variables we think are worth looking at because if we put too many in, they’ll get confounded and…we’ll just find nonsense.
“In the machine learning type of world, of course what we’re doing is we are trying to throw in as many we reasonably can that are sensible variables to use and let the algorithms help us decide which ones are worth keeping. But the humans are essential.”
Watch the preview of the panel discussion, which is now on demand: