CS–Assignment #2 Solution

Description

Note 1: Your submission header must have the format as shown in the above-enclosed rounded rectangle.

Note 2: Homework is to be done individually. You may discuss the homework problems with your fellow students, but you are NOT allowed to copy – either in part or in whole – anyone else’s answers.

Note 3: Your deliverable should be a .pdf file submitted through Gradescope until the deadline. Do not forget to assign a page to each of your answers when making a submission. In addition, source code (.py files) should be added to an online repository (e.g., github) to be downloaded and executed later.

Note 4: All submitted materials must be legible. Figures/diagrams must have good quality.

Note 5: Please use and check the Canvas discussion for further instructions, questions, answers, and hints.

1. [16 points] Considering that ID3 built the decision tree below after analyzing a given training set, answer the following questions:

Tear

^Normal Reduced

2. [15 points] Complete the Python program (decision_tree.py) that will read the files contact_lens_training_1.csv, contact_lens_training_2.csv, and contact_lens_training_3.csv. Each of those training sets has a different number of instances. You will observe that now the trees are being created setting the parameter max_depth = 3, which it is used to define the maximum depth of the tree (pre-pruning strategy) in sklearn. Your goal is to train, test, and output the performance of the models created by using each training set on the test set provided (contact_lens_test.csv). You must repeat this process 10 times (train and test by using a different training set), choosing the lowest accuracy as the final classification performance of each model.

2. [32 points] Consider the dataset below to answer the following questions:

y

x

1. [4 points] What is the leave-one-out cross-validation error rate (LOO-CV) for 1NN? Use Euclidean distance as your distance measure and the error rate calculated as:

⁼

1. 2. [4 points] What is the leave-one-out cross-validation error rate (LOO-CV) for 3NN?

1. 2. [4 points] What is the leave-one-out cross-validation error rate (LOO-CV) for 9NN?

1. 2. [5 points] Draw de decision boundary learned by the 1NN algorithm.

1. 2. [15 points] Complete the Python program (knn.py) that will read the file binary_points.csv and output the LOO-CV error rate for 1NN (same answer of part a).

4. [12 points] Find the class of instance #10 below following the 3NN strategy. Use Euclidean distance as your distance measure. You must show all your calculations for full credit.

5. [25 points] Use the dataset below to answer the next questions:

1. [10 points] Classify the instance ‹D15, Sunny, Mild, Normal, Weak› following the Naïve Bayes strategy. Show all your calculations until the final normalized probability values.

2. [15 points] Complete the Python program (naïve_bayes.py) that will read the file weather_training.csv (training set) and output the classification of each test instance from the file weather_test (test set) if the classification confidence is >= 0.75. Sample of output:

Day	Outlook Temperature		Humidity	Wind	PlayTennis	Confidence
D15	Sunny	Hot	High	Weak	No	0.86
D16	Sunny	Mild	High	Weak	Yes	0.78

Important Note: Answers to all questions should be written clearly, concisely, and unmistakably delineated. You may resubmit multiple times until the deadline (the last submission will be considered).

NO LATE ASSIGNMENTS WILL BE ACCEPTED. ALWAYS SUBMIT WHATEVER YOU HAVE COMPLETED FOR PARTIAL CREDIT BEFORE THE DEADLINE!