Contributed Paper C185
Comparing Machine Learning algorithms with regression models in the classroom
AuthorsSarawuth Chesoh (Thailand)
Potjamas Chuangchang (Thailand)
Nittaya McNeil (Thailand)
PresenterSarawuth Chesoh (Thailand)
Despite arguments by Leo Breiman (2001) advocating the use of Machine Learning algorithms such as random forests and neural networks instead of classical regression models for prediction, many statisticians still prefer regression models. David Donoho (2015) suggested that these issues are best settled by doing studies that apply reproducible computer programs to readily available data sets. In this paper we present several examples involving different data types and sample sizes that show how this can be done effectively in the classroom by students using Monte Carlo simulation to generate data sets and web-downloadable computer programs to compare methods. The key element is that the students can easily and painlessly learn to do these studies themselves using their own computers both in and outside the classroom.