From the course: Machine Learning and AI Foundations: Classification Modeling

Overview

From the course: Machine Learning and AI Foundations: Classification Modeling

Start my 1-month free trial

Overview

“

- [Instructor] Okay, we're about to discuss about a dozen different algorithms for binary classification. Now, in the section after our discussion of the algorithms, we're gonna talk about some common problems. I want you to be aware of what these problems are so that you can be listening for these issues as you learn about the algorithms. In fact, you may want to watch the algorithms section now, then see the section on the common problems, and then return and listen to the lectures on the algorithms a second time. So here's some of the things that we'll be seeing in the next section that you should be listening for in this section. I won't always make explicit reference to them, but the algorithms will overlap with these issues. One is whether or not the variables contribute to the model individually or if they also interact. Another will be how it handles missing data. We frequently have missing data, and these algorithms will handle that all in a different way. Also, overfitting, we want our models to be just complex enough. Not complex enough, and we have poor accuracy. Too complex, and then it doesn't fit new data well. Finally, feature selection, how does it go about determining which variables to use and how important those variables will be? Now again, I won't always make explicit reference to each of these issues when we discuss each algorithm. Focus on what the algorithms are doing conceptually, but here are some specific things to attend to. Does it use all or just some of the input variables? Does it use all or just some of the case data? When you're listening to me describe the algorithms, reflect on situations where they might perform well and situations where they might perform poorly.

Contents