See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/370591757

Question bank on decision tree algorithm

Presentation · May 2023

DOI: 10.13140/RG.2.2.12228.09606

CITATIONS READS

0 6,090

1 author:

Abhishek D. Patange
ABB
117 PUBLICATIONS 946 CITATIONS

SEE PROFILE

All content following this page was uploaded by Abhishek D. Patange on 07 May 2023.

The user has requested enhancement of the downloaded file.

 Question bank on decision tree algorithm

DECISION TREE – ENTROPY & INFORMATION GAIN

Problems for 2-4 marks

Problems 1
A decision tree classifier is used to predict the failure mode of a boiler system with a dataset
of 300 samples, where 200 samples indicate normal operation and 100 samples indicate tube
leakage. What is the entropy of the classifier?

Problems 2
If a decision tree classifier is used to predict the type of heat exchanger fouling in a system
with a dataset of 180 samples, where 100 samples indicate deposition and 80 samples
indicate scaling, what is the entropy of the classifier?

Problems 3
We want to predict the type of tire wear in a vehicle with a dataset of 120 samples, where 50
samples indicate wear due to underinflation, 30 samples indicate wear due to overinflation,
and 40 samples indicate wear due to misalignment, what is the entropy of the classifier?

Problems 4
A dataset of 250 samples were collected to predict the type of suspension failure in a vehicle,
where 100 samples indicate failure due to worn-out shock absorbers, 80 samples indicate
failure due to broken springs, and 70 samples indicate failure due to worn-out bushings. What
is the entropy of the classifier?

Problems 5
We need to split a set of energy efficiency data on the attribute "glazing area distribution"
with two possible values (0.4 and 0.7) and the information gain is 0.214. What is the entropy
of the original set?

Problems 6
The information gain for splitting a set of solar panel efficiency data on the attribute "angle of
incidence" with three possible values (0°, 45°, and 90°) is 0.325. If the entropy of the original
set is 0.95, what is the size of the original set?

Problems 7
A decision tree classifier is trained on a set of geothermal energy data with two attributes:
"temperature" and "depth." The entropy of the original set is 0.81, and the information gain
for splitting on the "depth" attribute is 0.251. What is the information gain for splitting on the
"temperature" attribute?

Problems 8
Suppose a manufacturer of mechanical components is testing a new production process that
produces components with two possible defects: crack or deformation. The manufacturer has
collected data on 500 components produced by the new process, and has labeled each
component as either defect-free, cracked, or deformed. The data is summarized in the
following table:
Component status Defect-free Cracked Deformed
Number of components 350 100 50
What is the entropy of the dataset?

Dr. Abhishek D. Patange, Mechanical Engineering, COEP Tech, Pune Page 1 of 8

 Question bank on decision tree algorithm

Problems 9
A car manufacturer is testing a new engine design that can produce three types of failures:
overheating, oil leakage, and poor fuel efficiency. A data has been collected on 500 engines
produced by the new design, and has labeled each engine as either failure-free, overheating,
oil leakage, or poor fuel efficiency. The data is summarized in the following table:
Engine status Number of engines
Failure-free 350
Overheating 100
Oil leakage 30
Poor fuel efficiency 20
What is the entropy of the dataset?

Problems for 6-8 marks

Problems 10
A training dataset collected while navigating autonomous vehicle is shown below.
Speed Distance to Traffic Control Pedestrian Presence Decision
intersection Device
0.5 0.7 Stop sign Pedestrian crossing Slow down
0.2 0.9 Green traffic light No pedestrian crossing Stop
0.8 0.4 Red traffic light No pedestrian crossing Go
0.6 0.1 Pedestrian crossing No traffic light Slow down
0.3 0.5 No traffic sign Pedestrian crossing Go
0.9 0.8 No traffic sign No pedestrian crossing Stop
0.7 0.2 Green traffic light No pedestrian crossing Stop
0.4 0.6 Stop sign No pedestrian crossing Slow down
0.1 0.3 No traffic sign No pedestrian crossing Stop
0.5 0.5 Red traffic light Pedestrian crossing Go
Identify discrete & continuous attributes along with their features. Calculate information gain
of two discrete attributes individually.

Problems 11
A decision tree classifier is to be trained to predict the likelihood of a vehicle rolling over
during a sharp turn based on input features: vehicle speed during the turn (mph), angle of the
turn (degrees), type of vehicle (e.g., sedan, SUV, pickup truck), weight (Ibs), road surface
conditions (e.g., dry, wet, icy), tire tread, driver behavior (e.g., aggressive, cautious).
 Calculate information gain of attribute ‘Vehicle Type’ & ‘Road Surface’.
 Compare them and comment on which one of these two is suitable for the best split.
Speed Turn Vehicle Weight Road Tire Driver
Rollover
(mph) Angle Type (lbs) Surface Tread Behavior
35 50° SUV 4000 Wet Good Aggressive Yes
25 30° Sedan 3200 Dry Good Cautious No
40 70° Pickup 5000 Icy Poor Aggressive Yes
30 45° SUV 4500 Wet Good Cautious No
50 80° Pickup 5500 Dry Good Aggressive Yes
20 20° Sedan 2800 Dry Good Cautious No
30 60° SUV 4200 Wet Good Aggressive Yes
40 75° Pickup 5100 Icy Poor Cautious Yes
35 55° SUV 4100 Wet Good Cautious No
45 90° Pickup 5800 Dry Good Aggressive Yes

Dr. Abhishek D. Patange, Mechanical Engineering, COEP Tech, Pune Page 2 of 8

 Question bank on decision tree algorithm

Problems 12
Consider following dataset
Roll Pitch Gyro Mode Gyro Speed Label
20.5 45.2 Stabilization mode Low speed Roll right
-15 60 Acrobatic mode High speed Roll left
10 30 Follow mode Low speed Pitch up
0 0 Stabilization mode High speed Pitch down
45 80 Acrobatic mode Low speed Roll right
-30 -60 Follow mode High speed Roll left
-10 -20 Stabilization mode Low speed Pitch down
60 -45 Acrobatic mode High speed Pitch up
-80 0 Follow mode Low speed Roll left
35 -70 Stabilization mode High speed Pitch down
Calculate entropy of dataset before splitting any attribute and after splitting Gyro Mode &
Gyro Speed.

Problems 13
Calculate entropy of dataset before splitting any attribute and after air filter type.
Room Size Number of Time of Day Air Filter Type Label
(sqm) Occupants (hours)
20 2 14 Basic Filter Comfortable
25 4 12 HEPA Filter Comfortable
30 6 16 Basic Filter Too Hot
18 3 10 Carbon Filter Comfortable
15 1 18 HEPA Filter Too Cold
28 5 13 Carbon Filter Comfortable
22 2 15 Basic Filter Comfortable
16 2 11 HEPA Filter Too Hot
21 4 14 Carbon Filter Comfortable
24 3 17 Basic Filter Too Cold

Problems 14
A government agency wants to investigate the factors that contribute to unsafe braking in cars
equipped with an ABS system. They have collected data on vehicle speed, road surface type,
brake pedal pressure, and ABS activation type for a set of cars involved in accidents where
the ABS system failed to prevent unsafe braking. Built a decision tree based on attribute
‘ABS Activation Type’ only.
Vehicle Speed Road Surface Brake Pedal Activation Label
(km/h) Type Pressure (%) Type
60 Dry 50 Active Safe
80 Wet 60 Passive Unsafe
70 Dry 70 Active Safe
50 Wet 30 Passive Unsafe
100 Dry 80 Active Safe
40 Wet 40 Passive Unsafe
90 Dry 90 Active Safe
65 Wet 20 Passive Unsafe
75 Dry 60 Active Safe
55 Wet 50 Passive Unsafe

Dr. Abhishek D. Patange, Mechanical Engineering, COEP Tech, Pune Page 3 of 8

 Question bank on decision tree algorithm

Problems 15
Consider following dataset
Road Suspension Speed Temperature Performance
Condition Type (mph) (°F)
Wet Independent 50 70 Good
Dry Dependent 70 80 Good
Wet Independent 30 60 Poor
Dry Dependent 60 90 Good
Wet Independent 40 75 Poor
Dry Dependent 80 85 Good
Wet Independent 20 50 Poor
Dry Dependent 50 70 Poor
Wet Independent 60 80 Good
Dry Dependent 40 65 Poor
In order to create a decision tree that accurately predicts the suspension system's
performance, which attribute is the most important amongst Road Condition and Suspension
Type?

Problems 16
Predict the Powertrain Output based on the Road Condition and Traction Control attributes.
Engine Throttle Road Traction Powertrain
Speed Position Condition Control Output
(rpm) (%)
2000 30 Dry Off Low
2500 50 Wet On Low
3500 80 Dry Off Medium
4000 90 Snowy On Low
3000 70 Wet On Medium
1500 20 Dry Off Low
4000 100 Dry Off High
2000 40 Snowy On Low
3000 60 Wet Off Medium
3500 80 Dry On High

Problems 17
Given the dataset, can you classify TATA car models as "Good", "Average", or "Poor" based
on their overall performance? What features are the most important for making this
classification?
Model Transmission Fuel Type Drivetrain Car Type Outcome
Tiago Manual Petrol Front-wheel drive Hatchback Good
Tigor Manual Petrol Front-wheel drive Sedan Average
Nexon Automatic Diesel All-wheel drive SUV Good
Harrier Automatic Diesel Front-wheel drive SUV Good
Altroz Manual Petrol Front-wheel drive Hatchback Good
Safari Automatic Diesel All-wheel drive SUV Average
Hexa Automatic Diesel All-wheel drive SUV Average
Bolt Manual Petrol Front-wheel drive Hatchback Average
Zest Manual Diesel Front-wheel drive Sedan Average
Sumo Manual Diesel Rear-wheel drive SUV Poor

Dr. Abhishek D. Patange, Mechanical Engineering, COEP Tech, Pune Page 4 of 8

 Question bank on decision tree algorithm

Problems 18
Calculate the entropy of the dataset for each attribute (altitude, wind, temperature, and
humidity) and determine which attribute is the best choice for the root node of the decision
tree.
Altitude Wind Temperature Humidity Outcome
High Low Hot High Crash
Low High Cold Low Safe
Low Low Mild High Safe
Medium High Hot Low Crash
High Low Mild Low Safe
Medium High Mild High Crash
High Low Cold High Crash
Low Low Cold Low Safe
Medium Low Mild Low Safe
Low High Hot High Crash

Problems 19
Given the training dataset, in order to build a decision tree to determine the optimal
destination for a drone which attribute amongst altitude, speed, wind, temperature, and
weather conditions.is most significant?
Altitude Speed Wind Temperature Weather Destination
High Fast Weak Warm Sunny City
Low Slow Strong Cold Rainy Forest
Medium Medium Weak Mild Cloudy Beach
High Slow Strong Warm Cloudy City
Medium Fast Weak Hot Sunny Beach
Low Medium Strong Cold Rainy Forest
High Slow Weak Warm Cloudy City
Low Fast Strong Hot Sunny Forest
Medium Medium Weak Mild Cloudy Beach
Low Slow Strong Cold Rainy Forest

Problems 20
In this dataset, we have five discrete attributes: Object Shape, Object Size, Object Weight,
Object Color, and Target Location, and the Target Location is the target variable we want to
predict.
Object Shape Object Size Object Weight Object Color Target Location
Square Small Light Red Shelf 1
Circle Medium Heavy Blue Shelf 2
Rectangle Small Light Green Shelf 3
Triangle Large Heavy Red Shelf 4
Circle Small Light Blue Shelf 2
Square Medium Heavy Green Shelf 3
Rectangle Small Heavy Red Shelf 4
Triangle Large Light Blue Shelf 2
Circle Small Heavy Green Shelf 3
Square Medium Light Red Shelf 1
 What is the entropy of the Target Location attribute in the entire dataset?
 Given the Object Shape attribute, what is the entropy of the Target Location attribute?
Dr. Abhishek D. Patange, Mechanical Engineering, COEP Tech, Pune Page 5 of 8
 Question bank on decision tree algorithm

 Given the Object Size attribute, what is the entropy of the Target Location attribute?
 Given the Object Weight attribute, what is the entropy of the Target Location
attribute?
 Given the Object Color attribute, what is the entropy of the Target Location attribute?
 What is the best attribute to split the dataset on to maximize information gain?

Problems 21
What is the entropy of the target variable (successful landing or not)?
Which attribute in the dataset has the highest information gain for predicting the target
variable?
Terrain Crater Sunlight Gravity Obstacles Landing Dust Communication Successful
Type Depth Site Level Landing
Rocky Shallow Strong Low None Near Low Good Y
Equator
Sandy Deep Weak High Few Near High Poor N
Poles
Crater Shallow Strong Low None Near Low Good Y
Equator
Rocky Deep Weak High Many Near High Poor N
Poles
Flat Shallow Strong Low None Near Low Good Y
Equator
Sandy Deep Strong High Few Near High Poor N
Poles
Crater Shallow Weak Low None Near Low Good Y
Equator
Rocky Deep Strong High Many Near High Poor N
Poles
Flat Shallow Weak Low None Near Low Good Y
Equator
Sandy Deep Strong High Few Near High Poor N
Poles

Problems 22
A company is trying to develop a decision tree classifier to predict whether a customer's
HVAC system needs a repair or not based on several features. The features are: outside
temperature (in degrees Fahrenheit), inside temperature (in degrees Fahrenheit), humidity
level (in percentage), and age of the HVAC system (in years). If the company splits the data
on the "outside temperature" feature and calculates the information gain, what is the
information gain value?
Outside Temp (F) Inside Temp (F) Humidity (%) Age (Years) Repair Needed?
<=60 <=62 <=40 <=5 No
>80 >75 >55 >5 and <=10 Yes
>80 >75 >55 >5 and <=10 Yes
>60 and <=70 >62 and <=68 <=40 <=5 No
>70 and <=80 >68 and <=75 >40 and <=55 <=5 No
>80 >75 >55 <=5 Yes
<=60 >62 and <=68 <=40 >5 and <=10 No
>80 >68 and <=75 >40 and <=55 >5 and <=10 Yes
>70 and <=80 >62 and <=68 >40 and <=55 <=5 No
>80 >75 >55 >10 Yes

Dr. Abhishek D. Patange, Mechanical Engineering, COEP Tech, Pune Page 6 of 8

 Question bank on decision tree algorithm

Tricky problems for 1-2 marks

23. In a decision tree, if a node has 3 possible outcomes with probabilities 0.4, 0.3, and 0.3,
what is its entropy?
24. Consider a decision tree with 5 levels and 32 leaves. How many nodes does it have?
25. What is the information gain if a binary split divides a dataset with 10 positive and 10
negative instances into two subsets, each with 5 positive and 5 negative instances?
26. In a decision tree, if a node has 5 possible outcomes with probabilities 0.1, 0.2, 0.3, 0.2,
and 0.2, what is its Gini index?
27. Consider a decision tree with 4 levels and 16 leaves. How many branches does it have?
28. In a decision tree, if a node has 4 possible outcomes with probabilities 0.25, 0.25, 0.25,
and 0.25, what is its information entropy?
29. What is the maximum possible value of the information gain for a binary split of a
dataset with 10 positive and 10 negative instances?
30. Consider a decision tree with 3 levels and 8 leaves. How many decision nodes does it
have?
31. In a decision tree, if a node has 6 possible outcomes with probabilities 0.1, 0.2, 0.1, 0.2,
0.2, and 0.2, what is its Gini index?
32. What is the minimum possible value of the information gain for a binary split of a dataset
with 10 instances, all of which belong to the same class?
33. A fluid has a density of 800 kg/m^3 and a viscosity of 0.05 Pa·s. It flows through a 3-cm
diameter pipe at a velocity of 5 m/s. What is the information gain of splitting the data
based on the flow rate, which can either be 10 L/min or 20 L/min?
34. A tank contains water up to a height of 2 m. The tank has a diameter of 3 m. What is the
information gain of splitting the data based on the tank material, which can either be steel
or concrete?
35. Suppose we have a dataset with 100 examples, 40 of which belong to class A and 60 of
which belong to class B. The dataset has one feature, and the feature has two possible
values. If 30 examples have the first value of the feature and 70 examples have the
second value, what is the information gain?
36. Suppose we have a dataset with 100 examples, and 80 of them belong to class A while
20 belong to class B. We split the data based on a feature that has 70 examples belonging
to A and 30 belonging to B in one branch, and 10 belonging to A and 10 belonging to B
in the other branch. What is the information gain of this split?
37. A decision tree is being developed to classify whether a bridge is safe or not based on
various factors. Out of 100 bridges, 60 are safe and 40 are unsafe. One of the factors
being considered is the material used to construct the bridge. Of the safe bridges, 30 are
made of steel and 30 are made of concrete. Of the unsafe bridges, 20 are made of steel
and 20 are made of concrete. Calculate the information gain for the material feature.
38. A materials engineer is developing a decision tree to classify different types of metals
based on their mechanical properties. The engineer has collected a dataset with 100
samples of metals, where 40 samples are aluminum, 30 samples are steel, and 30 samples
are copper. The engineer decides to split the dataset based on the tensile strength of the
metals, where samples with a tensile strength greater than 500 MPa are classified as
"strong" and those with less than 500 MPa are classified as "weak". Calculate the
information gain for this split.
39. Suppose we have a dataset of 1000 instances of car accidents with corresponding values
of accelerometer readings (in g units) and whether the airbag deployed or not. We split
the dataset based on the accelerometer reading being above or below 10g. The split
results in two subsets: 600 instances with readings below 10g (out of which 100 have
airbag deployed) and 400 instances with readings above 10g (out of which 300 have

Dr. Abhishek D. Patange, Mechanical Engineering, COEP Tech, Pune Page 7 of 8

 Question bank on decision tree algorithm

airbag deployed). What is the information gain of this split based on the airbag
deployment status?
40. Suppose we have a dataset of 100 accelerometers, out of which 60 are MEMS and 40 are
piezoelectric. We want to create a decision tree classifier based on two attributes:
frequency response and sensitivity. What are possible discrete (categorical)features to
these attributes?

Simple theory questions for 4-6 marks

41. What is the relationship between entropy and information in decision trees?
42. How is entropy used to measure the impurity of a decision tree node?
43. How does the information gain criterion help in selecting the best split in a decision tree?
44. Can decision trees handle continuous data? If so, how is entropy used to handle
continuous data in decision trees?
45. What is overfitting in decision trees, and how can it be avoided?
46. Can decision trees handle missing data? If so, how is entropy used to handle missing data
in decision trees?
47. What is the maximum possible entropy of a decision tree node, and when is it achieved?
48. Can decision trees be used for regression problems, and if so, how is entropy used in
regression decision trees?
49. How can decision trees be used for feature selection, and how does entropy play a role in
feature selection?
50. What are the limitations of decision trees, and how can entropy be used to overcome
these limitations?

Dr. Abhishek D. Patange, Mechanical Engineering, COEP Tech, Pune Page 8 of 8

View publication stats