Description
Problem: Implement a fixed-depth decision tree algorithm, that is, the input to the ID3
algorithmwill include the training data and maximum depth of the tree to be learned. The code
skeleton as well as data sets for this assignment can be found on Camino.
Data Sets: The MONK’s Problems were the basis of a first international comparison of learning
algorithms. The training and test files for the three problems are named monks-X.train and
monks-X.test. There are six attributes/features (columns 2–7), and binary class labels (column
1). See monks.names for more details.
Visualization: The code skeleton provided contains a function render_dot_file(), which can be
used to generate .png images of the trees learned by both scikit-learn and your code. See the
documentation for render_dot_file() for additional details on usage.
a. (Learning Curves, 15 points) For depth = 1, …, 10, learn decision trees and compute the
average training and test errors on each of the three MONK’s problems. Make three
plots, one for each of the MONK’s problem sets, plotting training and testing error curves
together for each problem, with tree depth on the x-axis and error on the y-axis.
Note: You need to write your own function to learn the tree and cannot use scikit-learn’s
DecisionTreeClassifier for this question.
b. (Weak Learners, 15 points) For monks-1, report the visualized learned decision tree and the
confusion matrix on the test set for depth = 1, 3, 5. You may use scikit-learns’s confusion
matrix() function [1].
Note: You need to write your own function to learn the tree and cannot use scikit-learn’s
DecisionTreeClassifier for this question.
c. (scikit-learn, 10 points) For monks-1, use scikit-learn’s DecisionTreeClassifier [2] to learn a
decision tree using criterion=’entropy’ for depth = 1, 3, 5. report the visualized learned
decision tree and the confusion matrix on the test set for depth = 1, 3, 5. You may use scikitlearn’s confusion matrix() function [1].
[1] https://scikit-learn.org/stable/modules/generated/sklearn.metrics.confusion_matrix.html
[2] https://scikit-learn.org/stable/modules/generated/sklearn.tree.DecisionTreeClassifier.html