The mlr Package

Now, we shall go into learning how the mlr package offers a complete framework to work with many machine learning models and problem. Often, in many ML projects, one has to manage an overwhelming amount of detailing around numerous experiments (also called trial-and-error iterations). Each experiment consists of many pieces of training using different machine learning algorithms, performance measures, hyperparameters, resampling techniques and predictions. Unless we do not systematically analyze the information obtained in each experiment, we will not be able to come out with the best combination of parameter values.

Another advantage of using the mlr package comes from its rich collection of machine learning algorithms from various packages. We do not have to install multiple packages for different implementation of the machine learning algorithm anymore. Instead, mlr offers everything in one place. To understand this better, refer to the following table:

Figure 9.3: The mlr...