Chapter 4 Methods Used
A number of analytical methods are available for use such as decision-trees, classification trees, linear-regression, random forests, k-nearest neighbors, and others. Not all of these techniques makes sense for our purposes as they are used to predict diffrent types of information.
The main goal of our analysis is to predict how risky a particular car is using the
symboling scale: [-3, -2, -1, 0, +1, +2, +3]. Since this output variable has class levels, and is not continuous, the appropriate analysis method is a classification tree. Note that if we were to not be using the
symboling scale, then a regression tree would be appropriate as it would have output a continuous risk variable. The below chart graphically summarizes this nicely.