Since we established that education and education-num represent the same information, why not use education-num, which is a numerical variable, and drop education attribute. I tried this actually and first I got the ‘Not Converged’ error so I had to increase the max_iter for LogisticRegression model. Why is that?
Later in Module 1 we will discuss the importance of scaling to obtain convergence. As in this case the correlation is very high, you can keep either of the variables and the results should be equivalent.