At first glance, the role of the Auto Classifier node seems to be limited to choosing an algorithm. In fact, it arguably allows one to avoid the choice of algorithm altogether in that it automatically chooses the top three and then creates an ensemble. However, model accuracy owes more to good data prep than algorithm choice. So what is one to make of Auto Classifier?
It is certainly true that, lacking the time or training for a proper modeling phase, the Auto Classifier would possibly do a better job than a data miner selecting a single method at random. It is worth noting, however, that some data mining experts have suggested that mastering a single method and its settings is often superior to attempting to use a host of algorithms without that mastery. Readers of an intermediate guide such as this one presumably have both the allocated time and the training to do a more complete job than just using the Auto Classifier on default settings. So how do...