Now that we have a decision tree constructed, we can use it for classification problems.
The classification method is a recursive one which parses the dictionary and halts (& returns) when a leaf node is encountered. In such problems, the vast chunk of processing time goes into building the decision tree.
It would be very inconvenient if the decision tree is to re-built every time we classify a problem.
It makes a lot of sense to serialise the tree, and we use the Python pickle module to do so. The code is available via Bitbucket