4. Suppose I design a naive classification algorithm as follows. If all inputs in my training data is +, I output a +, else I output a -. For a dataset with 50% + and 50% - classes, what is its leave-one-out cross validation error. (10 points)

5. Consider that there are 2 classes C1 and C2. Based on entropy, which feature is better to split on for a decision tree in the following case (the numbers indicate the number of instances of the specified class) (15 points)

5. Consider that there are 2 classes C1 and C2. Based on entropy, which feature is better to split on for a decision tree in the following case (the numbers indicate the number of instances of the specified class) (15 points) Feature-1 Feature-3 Feature-2 12 instances instances instances C1 C1 CO CO C1 CO C1 CO C1 CO C1 C0 2 2 CO C1 2 2 CO C1 5 5 3 1 CO C1 1 1 0 1 0 1 2