1)Load the iris sample dataset from sklearn (load_iris()) into Python using a Pandas dataframe. Induce a set of binary Decision Trees with a minimum of 2 instances in the leaves, no splits of subsets below 5, and an maximal tree depth from 1 to 5 (you can leave the majority parameter to 95%). Which depth values result in the highest Recall? Why? Which value resulted in the lowest Precision? Why? Which value results in the best F1 score? Explain the difference between the micro/macro/weighted methods of score calculation.

2)Simulate a binary classification dataset with a single feature using a mixture of normal distributions with NumPy (Hint: Generate two data frames with the random number and a class label, and combine them together). The normal distribution parameters (np.random.normal) should be (5,2) and (-5,2) for the pair of samples. Induce a binary Decision Tree of maximum depth 2, and obtain the threshold value for the feature in the first split. How does this value compare to the empirical distribution of the feature?