Catch defective parts from machine readings, where a straight line fails and an RBF kernel works.
New to this? See the worked example firstA support vector machine draws the widest possible boundary between two classes, and with a kernel it can draw curves. In these questions defective parts come from machines running too hot or too cold, so a straight line fails and the RBF kernel shines.
You also tune C and gamma, see what happens without scaling, and use the margin to decide which parts need a human check.
If you are new to it, read the worked example first, then solve the questions in order: the early ones explore the data and train a first model, the middle ones test it on rows it has not seen, and the last ones tune it and turn the result into a business decision.