Analytics for the Internet of Things (IoT)

Comparing different models to find the best fit using R

It is best practice to not only find the optimal version of a trained ML model but to compare multiple trained optimized ML algorithms. Each ML algorithm has its strengths and weaknesses. Choosing only one is limiting yourself. Select several ones that are likely to grow into good trained models given the particular problem you are solving. Then, compare them against each other, and let the best model win.

There are several ways to compare ML models against each other. ROC charts and AUC measures are two of the more popular methods. We will introduce each next.