(15 points) Consider a scenario in which you train a decision tree to classify credit card transactions into fraudulent (+) or legitimate (-). You have a dataset with 10,000 labeled transactions (5,000 + and 5,000 -). You randomly select 8,000 examples to build a decision tree with a minimum leaf size of 2 data points, then test your tree on the held-out 2,000 data points. You observe that your training error is 0.01 but your test error is 0.20. Explain why this might happen and the steps that you could take to improve the performance of the decision tree algorithm