2019 EXAMINATIONS

Part II

COMPUTING AND COMMUNICATIONS [2.5 Hours]

SCC.361 Artificial Intelligence

Candidates are asked to answer THREE questions from FOUR; each question is worth a total
of 25 marks.

Use a separate answer book for each question.

Simple calculators permitted

Question 1

1.a Describe the 5 levels of autonomy including the sub-levels for levels 4 and 5. Give an
example of how each level should be interpreted in terms of controlling a car.
[9 marks]

1.b Describe the procedure of the k-means clustering algorithm.

[8 marks]

1.c For the following two data samples:

x1={9; -7; 5; 13}
x2={16; -42; 58; 4}

i. calculate the range and mean;

[4 marks]
ii. perform normalisation;
[4 marks]

Total 25 Marks
Question 2

2.a Describe the three main types of machine learning methods and give an example for
each one of them.
[6 marks]

2.b Sketch and compare biological and computational neurons.

[5 marks]

2.c Describe and compare the two main types of fuzzy rule-based systems. Provide an
example for each one of them.
[6 marks]

2.d In the following Table the expected profit of three investment decisions made by an
investment banker under different conditions (future oil price changes) is given. Using
Maximax, Maximin, Lagrangian (equal probabilities), and Minimum Regret Decision
Making methods determine which investment is best and what it will bring to the
investment banker.

Future oil Expected Profit

price changes
Transport & Travel Manufacturing Agriculture
Decrease £900M £150M £500M
Stable £250M £300M £500M
Increase £80M £500M £400M

[8 marks]

Total 25 Marks
Question 3
3.a Describe three common types of chromosome encoding methods and give an example
for each of them.

[6 marks]

3.b Given below is an example genetic algorithm with an initial population consisting of six
chromosomes.

C1 1010101010 C2 1101101101
C3 0001000100 C4 0010010010
C5 0000000000 C6 1000000000

i. The fitness value is calculated as the number of 1s present in the chromosome.

Calculate the fitness of each chromosome in the initial population.
[3 marks]

ii. If the Roulette Wheel Selection algorithm is used as the selection method, is
there any chromosome that will never be selected? Which chromosome and
why?
[4 marks]

iii. In order to select two parents, the Roulette Wheel Selection algorithm is used.
Assume that two randomly generated numbers are 0.6 and 0.9. Fill in the
following table and specify which chromosomes are selected as parents.

Chromosome C1 C2 C3 C4 C5 C6
Fitness
Probability
Cumulative
Probability
[8 marks]

iv. In order to generate two new offsprings, the Uniform Crossover is used for
recombination of the selected parent chromosomes in (iii). Assume that
Uniform Crossover uses the following randomly generated numbers and the
swapping probability is 0.5. What are the generated offsprings?

0.3 0.2 0.8 0.1 0.1 0.9 0.4 0.2 0.7 0.2

[4 marks]

Total 25 Marks
Question 4
4.a Describe the Naïve Bayes classifier. How does the Naïve Bayes classifier estimate
probabilities from continuous attributes?
[4 marks]

4.b Take the dataset below that informs about decision making attributes to play tennis
outside for different days. The dataset consists of 14 examples with four attributes
(“Outlook”, “Temperature”, “Humidity”, and “Wind”) and a Binary outcome (“Play
Tennis”). If you build an optimal decision tree according to the Information Gain (IG)
score, which attribute would be chosen as the root node of the decision tree? In order
to answer this question, calculate the Information Gain (IG) score for “Temperature”
and “Humidity” attributes only and fill in the following table. The IG score of attribute
A is: IG(A) = Entropy before – AvgEntropy after choosing A. You need to show your
calculations, and you only need to do this for the root node of the decision tree.

Attribute “Outlook” “Temperature” “Humidity” “Wind”

Information Gain 0.246 ? ? 0.0478

[15 marks]

Question 4 continues on next page…

Question 4 continued.

4.c Which attribute would be chosen as the root node of the decision tree? Why?

[2 marks]

4.d Consider the following confusion matrix of a classification system with two possible
classes on 500 samples. Calculate accuracy, precision, recall, and F1-Score for this
classification system.

Predicted Class

Class=“Yes” Class=“No”

Actual Class Class=“Yes” 210 40

Class=“No” 60 190

[4 marks]

Total 25 Marks

---End of Paper---