Practical Predictive Analytics

Basic decision tree concepts

Decision trees algorithm are analogous to the game of 20 questions, or animal, vegetable, mineral. A basic decision tree algorithm will start at the root of the tree, representing all of the data, and continue to break each category into two separate categories that are based upon optimal methods, to be able to guess the best characteristics to identify each of the two categories. This is known as splitting a node. Then the algorithm will continue to split the data until it either cannot split any more, or stops in other ways based upon the control parameters.

For example, if we want to classify gender based upon the presence of osteoporosis, ischemic heart disease, and the average number of prescriptions, we ...