Question 1.

The result of a confusion matrix is as below:

Confusion Matrix Actual value

faulty(0) Good(1)

Prediction faulty(0) 50 3

Good(1) 5 42

What is the accuracy value?

Group of answer choices

92%

8%

50%

42%

Question 2.

A researcher has conducted a research to evaluate the relationship between age and the height of the students in a class. His results revealed the correlation of -1.0 between age and height. What is the conclusion of this research?

Group of answer choices

There is a strong negative relationship between age and height

There is a strong positive relationship between age and height

There is no relationship between age and height

There is a weak relationship between age and height

Question 3.

Which of the following statements is correct?

Group of answer choices

Multivariable and Multivariate regression are the same thing, i.e. each involve multiple features (independent variables) and one response (dependent variable).

There is no way to make a 2-way split on continuous data.

If we complete an analysis and find that there is statistically significant evidence of a relationship between a predictor variable and its corresponding response variable we can safely assert that we know the predictor caused the response.

Decision tree induction algorithms use what is called a greedy strategy but then may only find a local minimum.

Question 4:

Remember the dataset of alligators in Lecture 3 which was about the length and weight of several aligators in Florida. . The variable X is the length of aligator and the Y variable is the weight of them. A researcher decided to use decision tree and designed two steps: X4 . What is the name of this method of splitting?

Group of answer choices

Entropy classification

Gini index

Binary splitting

Multi-way splitting

Question 5:

Which one is NOT one of the disadvantages of decision tree?

Group of answer choices

In a complex model, it may result to an overfitting

Small variations in features may affect the trees

It is not useful when we have targets with more than 2 levels

If some of the target variables overcome the others, it may result in biased trees

Question 6:

What is the minimum value of Entropy in a decision tree and what does it mean?

Group of answer choices

0.5- Maximum impurity

0- Minimum impurity

0.5- Minimum impurity

0- Maximum impurity

Question 7:

Which one is NOT one of the gradient descent properties?

Group of answer choices

Works well even with larger number of instances

Needs many iterations

Need to choose an alpha (learning rate)

Need to compute transpose(X)*X

Question 8

In Gradient descent technique, we chose an alpha value (learning rate) in computation of parameters (theta zero and theta 1). What will happen if we assign a very small value to alpha?

Group of answer choices

The speed of the computations will be very high

The model computations may take a long time to converge

The model may never converge

There will be no need to iterate

Question 9:

Consider the following table which shows the satisfaction of customers from different stores:

Excellent Good Poor Total

Kmart 272 477 251 1000

Sears 315 457 228 1000

JCP 323 470 207 1000

Wards 391 404 205 1000

Total 1301 1808 891 4000

what is the probability that a customer find a store to be Good store if he/she is in JCP?

Group of answer choices

0.26

0.47

0.12

0.25

Question 10:

Which statement is NOT correct about Regression?

Group of answer choices

We don’t need to choose learning rate in Normal equation technique of regression

The goal in Gradient descent algorithm is to minimize the cost function

The cost function is the difference between the hypothesis and predicted output

The mathematics utilizing a cost function can be applied to either linear or logistic regression.

Question 11:

An artificial intelligence model is designed to predict a color of a shape with a unique color. During the training some pictures with three color (red,green and blue) are shown to machine and on each iteration, the correct color is given to machine. Now, this machine can predict red, green and blue colors. After training, a circle with color purple is shown to machine. What is the type of training and what will be the prediction of the model?

Group of answer choices

Supervised- red, blue or green

Unsupervised- red, blue or green

Unsupervised- purple

Supervised- purple

Question 12:

What type of learning is reflected in following figure and why?

Group of answer choices

Supervised because the class is known

Unsupervised because the target is unknown

Supervised because the learning algorithm is specified

Unsupervised because the training and testing sets are different

Question 13:

Which of the following examples is an example of unsupervised learning?

Group of answer choices

A face recognition in phone whereas your phone takes some pictures of you and next time it will only allow YOU to open the phone based on your face

A machine can analyze some x-rays and can predict if someone has cancer or not based on his/her x-ray.

Based on past information about spams, filtering out a new incoming email into Inbox (normal) or Junk folder (Spam)

NASA discovers new heavenly bodies and finds them different from previously known astronomical objects – stars, planets, asteroids, blackholes etc. (i.e. it has no knowledge about these new bodies) and categorize them the way it would like to (distance from Milky way, intensity, gravitational force, red/blue shift or whatever)

Qustion 14:

Which statement is NOT correct about SVM for a problem with 2 set of input features and a binary class of output?

Group of answer choices

SVM maximizes the margin between support vectors

Hyperplane is the line that separates two classes

SVM is a good approach only for smaller datasets

Support vectors are the features that are closer to the hyperplane

Question 15:

A scatter plot of two features for predicting three classes is shown below. What is the best machine learning model that we can fit on this information?

Linear regression

It is impossible to make any classification on this data even with low accuracy

SVM with RBF kernel

Linear SVM