3. To be done by hand. Consider the data: x Y z 1 1 good good bad Class label A A 1 0 0 1 good 0 0 B We want to develop a decision tree to predict the class label based on the features X, Y and Z. Based on the gini criterion, decide which feature you will use to do the initial split of nodes. (In other words, what split (based on X, Y or Z?) will result in the most reduction of gini impurity? Clearly show all your work