475994 (1) [Avatar] Offline
#1
Greetings! Thank you for his excellent book, which has helped me significantly.

In many of the algorithms, such as random forest, multiple regression, and logistic regression, it's easy to see which variables contribute the most to the solution to a problem.

Do KNN and Naive Bayes have similar ways to view variable importance? I suspect that KNN does not--it simply uses all the variables equally. I've done some searching and determined that Naive Bayes does have a way to figure variable importance, but that is with pre-packaged functions. The way that Naive Bayes is explained in the book, building the function from the ground up, doesn't provide an obvious way to figure out the most important variables in the solution.

Any help you could provide would be greatly appreciated. Thanks!