COMP3308/3608 Artificial Intelligence
Weeks 5 Tutorial exercises
Introduction to Machine Learning. K-Nearest Neighbor and 1R.
Exercise 1 (Homework). K-Nearest Neighbor with numeric attributes
Consider the dataset given below where years_experience is the only attribute and salary is the class. What will be the prediction for the new example years_experience=5 using 1-Nearest-Neighbor and 3- Nearest-Neighbor with Manhattan distance?
COMP3308/3608 Artificial Intelligence, s1 2021
years_experience
salary
1
4
low
2
9
medium
3
8
medium
4
15
high
5
20
high
6
7
medium
7
12
medium
8
23
medium
9
1
low
10
16
medium
Exercise 2. K-Nearest Neighbor with nominal and numeric attributes
The following dataset describes adults as short, medium and tall based on two attributes: gender and height in meters.
name
gender
height
class
Cristina
F
1.6
short
Jim
M
2
tall
Margaret
F
1.9
medium
Stephanie
F
1.88
short
Caitlin
F
1.6
short
David
M
1.7
short
William
M
2.2
tall
Stephen
M
2.1
tall
Debbie
F
1.8
medium
Todd
M
1.95
medium
What would be the prediction of 5-Nearest-Neighbor using Euclidian distance for Maria who is (F, 1.75)? Show your calculations. Do not apply normalization for this exercise (but note that in practical situations you need to do this for the numeric attribute!). In case of ties, make random selection.
Hint: gender is nominal attribute; see the lecture slides about how to calculate distance for nominal attributes.
1
Exercise 3. 1R algorithm
Consider the iPhone dataset given below. There are 4 nominal attributes (age, income, student, and credit_rating) and the class is buys_iPhone with 2 values: yes and no. Predict the class of the following new example using the 1R algorithm: age<=30, income=medium, student=yes, credit-rating=fair. Break ties arbitrary.
COMP3308/3608 Artificial Intelligence, s1 2021
age
income
student
credit_rating
buys_iPhone
<=30
high
no
fair
no
<=30
high
no
excellent
no
31..40
high
no
fair
yes
>40
medium
no
fair
yes
>40
low
yes
fair
yes
>40
low
yes
excellent
no
31..40
low
yes
excellent
yes
<=30
medium
no
fair
no
<=30
low
yes
fair
yes
>40
medium
yes
fair
yes
<=30
medium
yes
excellent
yes
31..40
medium
no
excellent
yes
31..40
high
yes
fair
yes
>40
medium
no
excellent
no
Dataset adapted from J. Han and M. Kamber, Data Mining, Concepts and Techniques, 2nd edition, 2006, Morgan Kaufmann.
Exercise 4. Using Weka
Weka is an open-source ML software written in Java and developed at the University of Waikato in New Zealand, for more information see http://www.cs.waikato.ac.nz/ml/weka/index.html. It is a solid and well-established software written by ML experts.
Install Weka on your own computer at home, it will be required for Assignment 2. Meanwhile we will use it during the tutorials to better understand the ML algorithms we study.
1. Get familiar with Weka (Explorer). Load the weather data with nominal attributes (weather.nominal.arff). The datasets are in \Program Files\Weka-3.8.
2. Choose ZeroR classifier.
3. Choose “Percentage split” as ‘Test option” for evaluation of the classifier: 66% training set, 33% testing set
ZeroR is a very simple ML algorithm (also called “straw man”). It works as follows:
• Nominal class (classification) – determine the most frequent class in the training data most_freq; classification: for every new example (i.e. not in the training data), return most_freq • Numeric class (regression) – determine the average class value in the training data av_value; classification: for every new example (i.e. not in the training data), return av-value
ZeroR is useful to determine the baseline accuracy when comparing with other ML algorithms.
4. Run the IBk algorithm (k-nearest neighbor, IB comes from Instance-Based learning; it is under “Lazy”) with k=1 and 3.
Was IB1 more accurate than IB3 on the test data? Does a higher k mean better accuracy? How does IB1 and IB3 compare with ZeroR in terms of accuracy on test set?
2
5. Consider IB1. Change “Percentage split” to “Use training data”(i.e. you will build the classifier using the training data and then test it on the same set). What is the accuracy? Why? Would you achieve the same accuracy if IB1 was IBk, for k different than 1?
6. Now load another dataset – iris.arff. Run and compare the same algorithms. Don’t forget to normalize the data (the attributes are numeric). The normalization feature is under Filters-Unsupervised-Attribute- Normalise.
Why we didn’t normalize the weather data when we run IBLk above on the weather data?
7. Select the K-star algorithm, another lazy algorithm, a modification of the standard nearest neighbor. Read its short description (right click on the algorithm’s name -> “Show properties’ -> “More”). Run it and compare the results (accuracy on test set using “Percentage split” again) with the 1 and 3 nearest neighbor algorithms and ZeroR.
8. Now run the 1R algorithm on both the iris and weather.nominal datasets using the same evaluation procedure (“Percentage split”). 1R is under “rules” in Weka and called “OneR”.
What are the rules that 1R formed? How many rules? How many attributes in each rule? Which are these attributes for the two datasets?
9. Compare the accuracy of ZeroR, OneR, 1R, IB1, IBk on the test set. Which was the most accurate classifier?
COMP3308/3608 Artificial Intelligence, s1 2021
3