[Type the document title]

V Semester Diploma Examination, February/ March 2023

Artificial Intelligence and Machine Learning (20CS51I)
SCHEME OF VALUATION
Section-I
1a. Summarize the challenges associated with Machine Learning 5M .
Any 5 relevant challenges can be considered 1*5=5M
1b. How AI Software Development life cycle differs from traditional software development? 5M
Any five differences 1*5=5M
1c. Perform the following operations on Car manufacturing company dataset auto-mpg.csv given below
using pandas. 10M
a) Read data from an existing file 2.5M
b) statistical details of dataset 2.5M
c) Get all cars with 8 cylinders 2.5M
d) Get the number of cars manufactured in each year. 2.5M
2a. Differentiate between supervised machine learning and Unsupervised machine learning 5M
Any five differences 1*5=5M
2b. Explore different sources of big data in machine learning 5M.
Writing any 5 sources 1*5=5M
2c. Ramesh decides to walk 10000 steps every day to combat the effect that lockdown has had on his
body’s agility, mobility, flexibility and strength. Consider the following data from fitness tracker over a
period of 10 days 10M
Code to add1000 steps to all the observations 5M
Code to find out the days on which Ramesh walked more than 7000 steps 5M
Section-II
3a. How to handle missing values in the dataset? Explain. 10M
List approaches to handle the missing values - 2M + Explanation of any four methods(2*4=8M)
3b. A dataset is given to you for creating machine learning model. What are the steps followed before using
the data for training the model? Elaborate each step. 10M
Listing of steps 2M+Explaination of any four steps (2*4=8M )
4a. A company wants to study iris dataset to make predictions. However, the data gathered is not clean for
analysis. The company requests you to write a python code to perform the following operations for data
driven competitive advantage (Assume dataset with missing values) 10M .
Check for missing values 5M + Replace missing values with mean value 5M

1
 [Type the document title]

4b. Statistical summary of Credit card dataset is as follows. Analyse and explain statistical metrics from
above summary 10M
Analysis of given summary 2M + Explanation of all statistical metrics 8M
Section-III
5a. A Machine learning model was built to classify patient as COVID +ve(1) or – ve(0). The confusion
matrix for the model is as shown below. Evaluate accuracy, precision, recall, specificity and F1-Score.
10M
Accuracy – writing formula 1M+calculation 1M=2M
Precision- writing formula 1M+calculation 1M =2M
Recall -writing formula 1M+calculation 1M =2M
Specificity- writing formula 1M+calculation 1M =2M
F1-Score -writing formula 1M+calculation 1M=2M
5b. Compare overfitting with underfitting 5M (1*5=5M)
5c. How to Choose the Right Number of Clusters in k-means clustering? Explain any one method
5M. (Listing methods 1M+Explanation 4M )

6a. Cluster the following eight points (with (x, y) representing locations) into three clusters:
A1(2, 10), A2(2, 5), A3(8, 4), A4(5, 8), A5(7, 5), A6(6, 4), A7(1, 2), A8(4, 9)
Initial cluster centres are: A1(2, 10), A4(5, 8) and A7(1, 2). The distance function between two points a =
(x1, y1) and b = (x2, y2) is defined as Ρ(a, b) = |x2 – x1| + |y2 – y1|.
Use K-Means Algorithm to find the three cluster centres after the first iteration 10M
Calculating the distance from each point to all centres -5M
Finding which point belongs to which cluster and form new clusters 3M
Calculate the new cluster centre 2M

6b. Compare classification algorithms with clustering algorithm 5M

Any five comparisons 1*5=5M
6c. K-means clustering with Euclidean distance suffer from the curse of dimensionality. Is the statement
true and why? 5M
True(1M)+Explanation 4M
Section IV
7a. N-grams are defined as the combination of N keywords together. Consider the given sentence: “The
greatest glory in living lies not in never falling, but in rising every time we fall”
a. Generate bi grams for the above sentence
b. Generate tri-grams for the above sentence

2
 [Type the document title]

1. Generate bi grams (1M import library +4M for generating and printing output)
2. Generate tri-grams (1M import library +4M for generating and printing output)

7b. Discuss importance of dimensionality reduction in machine learning 5M

Any five valid points1*5=5M
7c. Summarise different strategies of production deployment 5M
Listing of deployment strategies 1M+Explaination 4M
8a. Demonstrate Stemming and Lemmatization concepts with suitable examples 10M
Stemming - Explanation 2.5M +Example 2.5M
Lemmatization - Explanation 2.5M +Example 2.5M

8b. Discuss different techniques of cross validation 5M

Listing of techniques 1M+ explanation of any two 4M
8c. What are MLOps? brief different stages that are involved in the MLOps lifecycles 5M (Definition 1M+
Explanation of MLOps lifecycle, any four stages 4M)
Section V
9a. With a neat diagram explain components of Docker 10M
Diagram 2M+ Docker component explanation 8M
9b. Demonstrate Simple Linear Regression considering a dataset that has two variables: salary (dependent
variable) and experience (Independent variable) 10M
Importing lib 1M +Reading Dataset 1M + pre-processing 2M+Split, build model 5M +finding score
1M
10a. Summarize any two cloud deployment models 10M
Any two cloud deployment models 5*2=10M
10b. Discuss any five ethical challenges in AI 10M
Any five ethical challengesI 2*5=10M

3
 [Type the document title]

V Semester Diploma Examination, February/ March 2023

Artificial Intelligence and Machine Learning (20CS51I)
MODEL ANSWERS
Section-I

Q.1a Summarize the challenges associated with Machine Learning. 5M

Ans:
1. Inadequate Training Data and Poor quality of data: The major issue that comes while using machine
learning algorithms is the lack of quality as well as quantity of data. Although data plays a vital role in the
processing of machine learning algorithms, many data scientists claim that inadequate data, inaccurate data,
noisy data, and unclean data are extremely exhausting the machine learning algorithms. This leads to less
accuracy in classification and low-quality results. Hence, data quality can also be considered as a major
common problem while processing machine learning algorithms.
2. Non-representative training data: To make sure our training model is generalized well or not, we have
to ensure that sample training data must be representative of new cases that we need to generalize. The
training data must cover all cases that are already occurred as well as occurring.
3. Overfitting: Overfitting is one of the most common issues faced by Machine Learning engineers and
data scientists. Whenever a machine learning model is trained with a huge amount of data, it starts
capturing noise and inaccurate data into the training data set. It negatively affects the performance of the
model. The main reason behind overfitting is using non-linear methods used in machine learning
algorithms as they build non-realistic data models.
4. Underfitting: Underfitting is just the opposite of Overfitting. Whenever a machine learning model is
trained with fewer amounts of data, and as a result, it provides incomplete and inaccurate data and destroys
the accuracy of the machine learning model.
5. Monitoring and maintenance: As we know that generalized output data is mandatory for any machine
learning model; hence, regular monitoring and maintenance become compulsory for the same. Different
results for different actions require data change; hence editing of codes as well as resources for monitoring
them also become necessary.
6. Lack of skilled resources: Although Machine Learning and Artificial Intelligence are continuously
growing in the market, still these industries are fresher in comparison to others. The absence of skilled
resources in the form of manpower is also an issue. Hence, we need manpower having in-depth knowledge
of mathematics, science, and technologies for developing and managing scientific substances for machine
learning.
7. Deployment: Deploying ML models in production environments can be difficult, as it requires expertise
in both ML and software engineering.
8. Model Selection: There are many different ML algorithms to choose from, and selecting the right one
for a particular problem is challenging.

4
 [Type the document title]

9. Feature Engineering: Extracting useful features from raw data is a critical step in ML, but it can be
difficult to identify the most relevant features.
Q. 1. b How AI Software Development life cycle differs from traditional software development?
5M
Ans: AI software development life cycle differs from traditional software development
1. Data-Driven: AI software development is heavily dependent on data and requires large amounts of
high-quality data to train and test models. In traditional software development, data is often an
afterthought.
2. Experimentation and Iteration: AI software development often involves a lot of experimentation and
iteration, as different algorithms and approaches are tried and tested to see which ones work best.
Traditional software development is typically more linear and follows a specific plan or design.
3. Model Selection: In AI software development, selecting the right model for a particular problem is
critical and can be a time-consuming process. In traditional software development, the choice of algorithms
and techniques is often predetermined.
4. Model evaluation and performance: In AI software development, model performance is evaluated
using different metrics and techniques, such as accuracy, precision, recall, and F1 score. In traditional
software development, model evaluation is often based on functional requirements.
5. Deployment and Maintenance: AI software deployment and maintenance requires additional
considerations, such as retraining models over time and deploying them in production environments. In
traditional software development, deployment and maintenance are often simpler and more
straightforward.
6. Traditional programs: In traditional programs, a developer designs logic or algorithms to solve a
problem. The program applies this logic to input and computes the output. But in Machine Learning, a
model is built from the data, and that model is the logic.
1 c .Perform the following operations on Car manufacturing company dataset auto-mpg.csv given
below using pandas. 10M
a) Read data from an existing file
b) Statistical details of dataset
c) Get all cars with 8 cylinders
d) Get the number of cars manufactured in each year.

5
 [Type the document title]

Ans:
a. Read data from an existing file
import pandas as pd
data = pd.read_csv(r"C:\Users\BASAPPA SIR\Desktop\Dataset\tauto-mpg.csv")
df= pd.DataFrame(data)
display(data)
OUTPUT:

b. Statistical details of dataset

stats = data.describe()
display(stats)

OUTPUT:

c. Get all cars with 4 cylinders

four_cylinder_cars = data[data['num-of-cylinders'] == 'four']

print(four_cylinder_cars)
OUTPUT:

6
 [Type the document title]

d. Get the number of cars manufactured in each year.

cars_horsehp = data.groupby (‘horsepower’)[‘horsepower'].count()

Print (cars_horsehp)
print(four_cylinder_cars)
OUTPUT:

Any equivalent code can be considered

2a. Differentiate between supervised machine learning and Unsupervised machine learning 5M

Supervised Learning Unsupervised Learning

Supervised learning algorithms are trained using Unsupervised learning algorithms are trained using
labelled data. unlabelled data.

Supervised learning model takes direct feedback to Unsupervised learning model does not take any
check if it is predicting correct output or not. feedback.
Unsupervised learning model finds the hidden patterns
Supervised learning model predicts the output. in data.
In supervised learning, input data is provided to the In unsupervised learning, only input data is provided to
model along with the output. the model.
The goal of supervised learning is to train the model
so that it can predict the output when it is given new The goal of unsupervised learning is to find the hidden
data. patterns and useful insights from the unknown dataset.
Supervised learning needs supervision to train the Unsupervised learning does not need any supervision to
model. train the model.
Supervised learning can be categorized Unsupervised Learning can be classified
in Classification and Regression problems. in Clustering and Associations problems.
Supervised learning can be used for those cases Unsupervised learning can be used for those cases
where we know the input as well as corresponding where we have only input data and no corresponding
outputs. output data.
It includes various algorithms such as Linear
Regression, Logistic Regression, Support Vector
Machine, Multi-class Classification, Decision tree, It includes various algorithms such as Clustering, KNN,
etc. and Apriori algorithm.
Supervised learning model produces an accurate Unsupervised learning model may give less accurate
result. result as compared to supervised learning.
(Any 5 can be considered)
7
 [Type the document title]

2b. Explore different sources of big data in machine learning 5M

Although Big data originates from multiple sources, below are the most common ones.
 Data collected from social Media sites like Facebook, WhatsApp, Twitter, YouTube, Instagram, etc
 Sensor placed in various places of the city that gathers data on temperature, humidity, etc. A camera
placed beside the road gather information about traffic condition and creates data. Security cameras
placed in sensitive areas like airports, railway stations, and shopping malls create a lot of data.
 IoT Appliance: Electronic devices that are connected to the internet create data for their smart
functionality, examples are a smart TV, smart washing machine, smart coffee machine, smart AC,
etc. It is machine-generated data that are created by sensors kept in various devices
 Customer feedback on the product or service of the various company on their website creates data.
For Example, retail commercial sites like Amazon, Walmart, Flipkart, and Myntra gather customer
feedback on the quality of their product and delivery time.
 E-commerce: In e-commerce transactions, business transactions, banking, and the stock market, lots
of records stored are considered one of the sources of big data. Payments through credit cards, debit
cards, or other electronic ways, all are kept recorded as data.
 Global Positioning System (GPS): GPS in the vehicle helps in monitoring the movement of the
vehicle to shorten the path to a destination to cut fuel, and time consumption. This system creates
huge data on vehicle position and movement.
 Transactional Data: Transactional data, as the name implies, is information obtained through online
and offline transactions at various points of sale. The data contains important information about
transactions, such as the date and time of the transaction, the location where it took place etc
2c. Ramesh decides to walk 10000 steps every day to combat the effect that lockdown has had on his
body’s agility, mobility, flexibility and strength. Consider the following data from fitness tracker
over a period of 10 days
10M

a. Perform an appropriate operation to add 1000 steps to all the observations

b. Find out the days on which he walked more than 7000 steps

8
 [Type the document title]

a. Perform an appropriate operation to add 1000 steps to all the observations

print(four_cylinder_cars)

b. Find out the days on which he walked more than 7000 steps

1a)

1b)

Any other equivalent code can be considered

Section-II
3a.How to handle missing values in the dataset? Explain. 10M Different
approachesto handle the missing valuesare as follows

1. Keepthemissingvalueasis
2. Removedataobjectswithmissingvalues(Deletingtheentirecolumn)
3. Removetheattributeswithmissingvalues(Deletingtheentirerow)
4. Estimateandimputemissingvalues

1. Keepthemissingvalueasis
Sometimes missing data is very less number of rows (say less than 3%) then we can
simply ignore themissingdata.Thereisnohardruleto keepthemissingdata itdepends onus
2. Removedataobjectswithmissingvalues(Deletingtheentirecolumn)
If a certain column has many missing values, then you can choose to drop the entire
column.Codetodroptheentirecolumnis asfollows:
df=train_df.drop(['Dependents'],axis=1)
df.isnull().sum()
3. Removetheattributeswithmissingvalues(Deletingtheentirerow)
Ifarowhasmanymissingvalues,thenyoucanchooseto dropthe entirerow.

9
 [Type the document title]

Codeto droptheentirerowisasfollows:
df =train_df.dropna(axis=0)
df.isnull().sum()

4. Estimateandimputemissingvalues
 ReplacingwithArbitraryValue
If you can make an educated guess about the missing value, then youcan replace it with
somearbitraryvalueusingthefollowingcode.
Ex.Inthefollowingcode,we are replacingthemissingvaluesofthe ‘Dependents’columnwith‘0’
train_df['Dependents']=train_df['Dependents'].fillna(0)
 ReplacingwithMean
This is the most common method of imputing missing values of numeric columns. One can use the
‘fillna’ method for imputing the columns ‘LoanAmount’ with the meanofthe
respectivecolumnvalues as below
train_df['LoanAmount'].fillna(train_df['LoanAmount'].mean())
 ReplacingwithMode
Modeisthemost frequently occurringvalue. Itisusedin thecaseofcategoricalfeatures.
You can use the ‘fillna’ method for imputing the categorical columns ‘Gender’,
‘Married’, and‘Self_Employed’.
train_df['Gender'].fillna(train_df['Gender'].mode()[0])
 ReplacingwithMedian
Median is the middlemost value. It’s better to use the median value for imputation in the case
ofoutliers. You can use ‘fillna’ method for imputing the column ‘Loan_Amt’ with the medianvalue.
train_df['Loan_Amt']=train_df['Loan_Amt'].fillna(train_df['Loan_Amt'].median()[0])
Note: Code is not mandatory for any approach

3b. A dataset is given to you for creating machine learning model. What are the steps followed before
using the data for training the model? Elaborate each step. 10M

1. Data Exploration: The first step is to explore the data and understand the characteristics of the dataset.
This includes understanding the number of observations and variables, the data types of each variable, and
the distribution of the data. This can be done by using summary statistics and visualizations such as
histograms, box plots, and scatter plots.
2. Data Cleaning: The next step is to clean the data. This includes handling missing or corrupted data,
removing outliers, and addressing any other data quality issues. This step is important because dirty data
can lead to inaccurate or unreliable models.
3. Data Transformation: After cleaning the data, it may be necessary to transform the data to make it
suitable for the machine learning model. This can include normalizing the data, scaling the data, or creating
new variables.
4. Feature Selection: Once the data is cleaned and transformed, it is important to select the relevant features
that will be used to train the model. This step can be done by using techniques such as correlation analysis,
principal component analysis, or mutual information.
5. Data Splitting: The next step is to split the data into training, validation, and test sets. The training set is
used to train the model, the validation set is used to tune the model's parameters, and the test set is used to
evaluate the model's performance.
6. Feature Engineering: This step is to create new features that will be useful in the model. This can include
creating interaction terms, polynomial terms, or binning variables.
7. Evaluation Metric: Selecting the right evaluation metric will help to evaluate the model's performance.
Common evaluation metrics include accuracy, precision, recall, F1 score, and area under the ROC curve.

10
 [Type the document title]

8. Model Selection: After the data is prepared, the next step is to select the appropriate machine learning
model. This can be done by comparing the performance of different models using the evaluation metric.
.

4a.A company wants to study iris dataset to make predictions. However, the data gathered is not
clean for analysis. The company requests you to write a python code to perform the following
operations for data driven competitive advantage(Assume dataset with missing values)
10M
 Check for missing values
 Replace missing values with mean value

Using isna() function to detect the missing values in a dataframe

Or
one can use isnull method to check missing values as below

Replacing missing values with mean value

This is the most common method of imputing missing values of numeric columns. You can use the ‘fillna’
method for imputing the numerical columns.
Suppose SepalLengthCm and PetalWidthCm columns are having missing values then below code can be
written to replace null values with mean using fillna method

11
 [Type the document title]

4b. Statistical summary of Credit card dataset is as follows. Analyse and explain statistical metrics
from above summary 10M

Analysis: The basic statistical summary provided appears to be describing a dataset

containing information about the credit card details of the customers, with columns “Age”,
“Annual Income” and Spending Score”. The summary tells age of credit card holder, his
annual income and spending score in the range 1-100 of 200 samples. These measures are
used to create a linear discriminate model to classify the species.
Explanation of statistical metrics
Count: The number of not-empty values for each variable and it is counted as 200 samples.
Mean: The average (mean) value of each variable and mean of age 38.85, annual income
60.56 and spending score average are 50.20.
Standard deviation: The standard deviation column gives a measure of how much the
values of each variable vary from the mean.
Min: The minimum value of age, annual income and spending score.
25%: It divides the data into four quartiles of 25%. Each shows the fourth quartile values of
a dataset.
50%: The 50% means the midpoint value of a dataset or it is the same as the second
quartile.
75%: The 75% means the percentage of splitting training dataset to train the model or the
same as the third quartile.
Max: The maximum value of age, annual income and spending score.

Section-III
5a. A Machine learning model was built to classify patient asCOVID +ve(1) or –ve(0). The
confusionmatrix for the model is as shown below. Evaluate accuracy, precision, recall, specificity
and
F1-Score. 10M

12
 [Type the document title]

TruePositive(TP):397
True Negative(TN) : 142
FalsePositive(FP):103
FalseNegative(FN): 126

5b. Compare overfitting with underfitting 5M

Overfitting Underfitting
Overfitting happens when we train a Underfitting occurs when the machine
machine learning model too much tuned learning model is not well-tuned to the
to the training set. training set.
As a result, the model learns the training The resulting model is not capturing the
data too well, but it can’t generate good relationship between input and output
predictions for unseen data. it produces well enough. Therefore, it doesn’t
low accuracy results for data points produce accurate predictions, even for the
unseen in training, hence, leads to non- training dataset. Resulting, an underfitting

13
 [Type the document title]

optimal decisions.Overfitting models model generates poor results that lead to

produce good predictions for data points high-error decisions.
in the training set but perform poorly on
It is not accurate for training set
new samples.
It is not accurate for validation set
It is accurate for training set
It is not accurate for validation set

Increase the number of features in the

Simplest technique you can use to
dataset to reduce underfitting
reduceOverfitting is Feature Selection

5c. How to Choose the Right Number of Clusters in k-means clustering? Explain any one method
5M
Three approaches to find the optimal number of clusters:
 The elbow method
 The optimization of the silhouette coefficient
 The gap statistic
Elbow method
1. Compute clustering algorithm (e.g., k-means clustering) for different values of k. For instance, by
varying k from 1 to 10 clusters.
2. For each k, calculate the total within-cluster sum of square (wss).
3. Plot the curve of wss according to the number of clusters k.
4. The location of a bend (knee) in the plot is generally considered as an indicator of the appropriate
number of clusters.

Average silhouette method

The algorithm is similar to the elbow method and can be computed as follow:
1. Compute clustering algorithm (e.g., k-means clustering) for different values of k. For instance, by
varying k from 1 to 10 clusters.
2. For each k, calculate the average silhouette of observations (avg.sil).
3. Plot the curve of avg.sil according to the number of clusters k.
4. The location of the maximum is considered as the appropriate number of clusters.

Gap statistic method

The algorithm works as follow:
1. Cluster the observed data, varying the number of clusters from k = 1, …, kmax, and compute the
corresponding total within intra-cluster variation Wk.

14
 [Type the document title]

2. Generate B reference data sets with a random uniform distribution. Cluster each of these reference
data sets with varying number of clusters k = 1, …, kmax, and compute the corresponding total
within intra-cluster variation Wkb.
3. Compute the estimated gap statistic as the deviation of the observed Wk value from its expected
value Wkb under the null hypothesis: Gap(k)=1B∑b=1Blog(W∗kb)−log(Wk)
4. Compute also the standard deviation of the statistics.
5. Choose the number of clusters as the smallest value of k such that the gap statistic is within one
standard deviation of the gap at k+1: Gap(k)≥Gap(k + 1)−sk + 1.

6a. Cluster the following eight points (with (x, y) representing locations) into three clusters:A1(2, 10),
A2(2, 5), A3(8, 4), A4(5, 8), A5(7, 5), A6(6, 4), A7(1, 2), A8(4, 9)
Initial cluster centres are: A1(2, 10), A4(5, 8) and A7(1, 2).The distance function between two points
a = (x1, y1) and b = (x2, y2) is defined as Ρ(a, b) = |x2 – x1| + |y2 – y1|.
Use K-Means Algorithm to find the three cluster centres after the first iteration10M
Iteration-01:
We calculate the distance of each point from each of the centre of the three clusters.
The distance is calculated by using the given distance function.
C1 : (2,10) C2 : (5,8) C3 : (1,2)
Calculating distance between A1(2,10) and C1,C2,C3
P(A1,C1) = |x2 – x1| + |y2 – y1|
= |2 -2 | + |10 - 10|
=0
P(A1,C2) = |x2 – x1| + |y2 – y1|
= | 5-2 | + | 8 - 10|
=5
P(A1, C3) = |x2 – x1| + |y2 – y1|
= | 1-2 | + | 2 - 10|
=9

Calculating distance between A2(2,5) and C1,C2,C3

P(A2,C1) = |x2 – x1| + |y2 – y1|
= |2 -2 | + |10 - 5|
=5
P(A2,C2) = |x2 – x1| + |y2 – y1|
= | 5-2 | + | 8 - 5|
=6
P(A2,C3) = |x2 – x1| + |y2 – y1|
= | 1-2 | + | 2 – 5 |
=4
Calculating distance between A3(8,4) and C1,C2,C3
P(A3,C1) = |x2 – x1| + |y2 – y1|
= | 2 -8 | + |10 - 4|
= 12
P(A3,C2) = |x2 – x1| + |y2 – y1|
= | 5-8 | + | 8 – 4 |
=7
P(A3,C3) = |x2 – x1| + |y2 – y1|
= | 1-8 | + | 2 – 4 |
=9
Calculating distance between A4(5,8) and C1,C2,C3
P(A4, C1) = |x2 – x1| + |y2 – y1|

15
 [Type the document title]

= | 2 -5 | + |10 -8 |
=5
P(A4, C2) = |x2 – x1| + |y2 – y1|
= | 5-5 | + | 8 -8 |
=0
P(A4, C3) = |x2 – x1| + |y2 – y1|
= | 1-5 | + | 2 –8 |
= 10

Calculating distance between A5(7,5) and C1,C2,C3

P(A5,C1) = |x2 – x1| + |y2 – y1|
= | 2 - 7 | + |10 - 5 |
= 10
P(A5,C2) = |x2 – x1| + |y2 – y1|
=|5-7|+|8-5|
=5
P(A5,C3) = |x2 – x1| + |y2 – y1|
=|1-7|+|2-5|
=9
Calculating distance between A6(6,4) and C1,C2,C3
P(A6,C1) = |x2 – x1| + |y2 – y1|
= | 2 - 6 | + |10 - 4 |
= 10
P(A6,C2) = |x2 – x1| + |y2 – y1|
=|5-6|+|8-4|
=5
P(A6,C3) = |x2 – x1| + |y2 – y1|
=|1-6|+|2-4|
=7
Calculating distance between A7(1,2) and C1,C2,C3
P(A7,C1) = |x2 – x1| + |y2 – y1|
= | 2 - 1 | + |10 - 2 |
=9
P(A7,C2) = |x2 – x1| + |y2 – y1|
=|5-1|+|8-2|
= 10
P(A7,C3) = |x2 – x1| + |y2 – y1|
= | 1 - 1 | + | 2 - 2 | =0
Calculating distance between A8(4,9) and C1,C2,C3
P(A8,C1) = |x2 – x1| + |y2 – y1|
= | 2 - 4 | + |10 - 9 |
=3
P(A8,C2) = |x2 – x1| + |y2 – y1|
=|5-4|+|8-9|
=2
P(A8,C3) = |x2 – x1| + |y2 – y1|
=|1-4|+|2-9|
= 10

We draw a table showing all the results.Using the table, we decide which point belongs to which
cluster.The given point belongs to that cluster whose centre is nearest to it.

16
 [Type the document title]

Distance from Distance from Distance from

Point belongs
Given Points centre (2, 10) of centre (5, 8) of centre (1, 2) of
to Cluster
Cluster-01 Cluster-02 Cluster-03
A1(2, 10) 0 5 9 C1
A2(2, 5) 5 6 4 C3
A3(8, 4) 12 7 9 C2

A4(5, 8) 5 0 10 C2
A5(7, 5) 10 5 9 C2
A6(6, 4) 10 5 7 C2
A7(1, 2) 9 10 0 C3
A8(4, 9) 3 2 10 C2

From here, New clusters are-

Cluster-01: First cluster contains points-A1(2, 10)

Cluster-02: Second cluster contains points-A3(8, 4),A4(5, 8),A5(7, 5),A6(6, 4),A8(4, 9)
Cluster-03: Third cluster contains points-A2(2, 5),A7(1, 2)
 We re-compute the new clusters.
 The new cluster centre is computed by taking mean of all the points contained in that cluster.
For Cluster-01:
We have only one point A1(2, 10) in Cluster-01, so, cluster centre remains the same.
For Cluster-02:
Canter of Cluster-02
= ((8 + 5 + 7 + 6 + 4)/5, (4 + 8 + 5 + 4 + 9)/5)
= (6, 6)
For Cluster-03:
Canter of Cluster-03
= ((2 + 1)/2, (5 + 2)/2)
= (1.5, 3.5)

This is completion of Iteration-01.

6b. Compare classification algorithms with clustering algorithm 5M

Classification Clustering
Classification is a supervised learning Clustering is an unsupervised learning
technique technique
Classification is used to predict a class or Clustering is used to group similar data
label for a given data point points together
classification algorithms include Decision Clustering algorithms include K-means,
Trees, Random Forest, Naive Bayes Hierarchical, DBSCAN
classification requires labelled data for Clustering does not require labelled data
training

17
 [Type the document title]

classification is mainly used for prediction Clustering is used to find pattern in data
classification is used to make predictions Clustering can be used to discover hidden
based on known patterns patterns or relationships in the data
Any other relevant points can also be considered

6c. K-means clustering with Euclidean distance suffer from the curse of dimensionality. Is the
statement true and why? 5M

Yes, the statement is true. The curse of dimensionality refers to the phenomenon that as the number of
dimensions (features) in a dataset increases, the amount of data needed to accurately estimate the
relationships between the points also increases exponentially. This can make it difficult to use
traditional statistical techniques, such as k-means clustering, which rely on calculating distances
between points.

In k-means clustering, the distance between points is typically measured using the Euclidean distance,
which is the straight-line distance between two points. As the number of dimensions increases, the
distance between points can become much larger, even if the points are relatively close to each other in
lower-dimensional space. This can cause k-means to become less effective at accurately identifying
clusters in the data.

One way to mitigate the curse of dimensionality in k-means clustering is to use dimensionality
reduction techniques, such as principal component analysis (PCA), to reduce the number of
dimensions in the dataset before applying k-means. Another option is to use a different clustering
algorithm, such as DBSCAN, which does not rely on calculating distances between points and is
therefore less affected by the curse of dimensionality.
Section IV

7a. N-grams are defined as the combination of N keywords together. Consider the given sentence:
“The greatest glory in living lies not in never falling, but in rising every time we fall”
a. Generate bi grams for the above sentence 10M
b. Generate tri-grams for the above sentence
a. bigram

b.trigram

18
 [Type the document title]

7b.Discuss importance of dimensionality reduction in machine learning 5M

The number of features or input variables that contribute to the prediction process to predict target or
output in machine learning model is called dimensionality. However, it is observed that all feature
variables do not have the same significance / contribution in the output.Therefore, before training a
machine learning model, it is necessary to identify relevant features that contribute to output / target and
this process is called dimensionality reduction
Dimensionality reduction is essential because of the following reasons:
1. To reduce complexity of model:More number of features / high dimensionality will lead to build a
complex model specially when there is a high correlation exits in the feature variables. Thus, it is
useful to select the right set of features to overcome this challenge.
2. To prevent overfitting:A dataset with high dimensionality sometimes may lead to overfitting of a
machine learning model because the model captures key features and noise as well. Thus, the model
performs well during training of the model, but performance degrades on testing on unknown data.
3. To achieve computational efficiency:A machine learning model with low dimensionality takes less
time in training of the model because it takes less time in computations.
4. It is useful to lower model training time and lower the data storage requirement.
5. It prevents from curse of dimensionality
7c. Summarise different strategies of production deployment 5M
Canary deployment releases the services incrementally in small phases, enabling the organizations to test
their applications against real users and simultaneously analyze their different versions for improvement.
Shadow deployment contains a new version and its old version alongside to test the predictions. A more
recent version is released after successful testing. A copy of the production environment traffic is sent to
the older version to ensure successful working.

Blue/green deployment maintains two stages in parallel: Blue for staging and green for production.
Quality assurance testing of new versions is carried out in the blue environment, and later, real traffic is

19
 [Type the document title]

sent from the green environment to the blue environment. Once the testing goes successful, the product is
shifted to the green environment.

Rolling deployment: This involves releasing a new version of an application to a small subset of servers
before releasing it to all servers. This allows for testing and monitoring before rolling out to the entire
server fleet

A/B testing: This involves releasing different versions of an application to different subsets of users in
order to test and compare the performance of each version.

8a. Demonstrate Stemming and Lemmatization concepts with suitable examples 10M

Lemmatization and Stemming are Text Normalization techniques. These techniques are used to prepare
words, text, and documents for further processing.
Stemming
It is the process of producing morphological variants of a root/base word. “boat” would be the stem for
[boat, boater, boating, boats].
Lemmatization is a method responsible for grouping different inflected forms of words into the root form,
having the same meaning. It is similar to stemming, in turn, it gives the stripped word that has some
dictionary meaning.
Lemmatization clearly identifies the base form of ‘troubled’ to ‘trouble’’ denoting some meaning whereas,
Stemming will cut out ‘ed’ part and convert it into ‘troubl’ which has the wrong meaning and spelling
errors.

‘troubled’ -> Lemmatization -> ‘troubled’, and error

‘troubled’ -> Stemming -> ‘troubl’
Ex for stemming

output

Ex for lemmatization

8b. Discuss different techniques of cross validation 5M

20
 [Type the document title]

1. Hold Out method

This is the simplest evaluation method and is widely used in Machine Learning projects. Here the entire
dataset(population) is divided into 2 sets – train set and test set. The data can be divided into 70-30 or 60-
40, 75-25 or 80-20, or even 50-50 depending on the use case. As a rule, the proportion of training data has
to be larger than the test data.

2. Leave One Out Cross-Validation

In this method, we divide the data into train and test sets – but with a twist. Instead of dividing the data into
2 subsets, we select a single observation as test data, and everything else is labelled as training data and the
model is trained. Now the 2nd observation is selected as test data and the model is trained on the remaining
data.

3. K-Fold Cross-Validation
In this resampling technique, the whole data is divided into k sets of almost equal sizes. The first set is
selected as the test set and the model is trained on the remaining k-1 sets. The test error rate is then
calculated after fitting the model to the test data.In the second iteration, the 2nd set is selected as a test set
and the remaining k-1 sets are used to train the data and the error is calculated. This process continues for
all the k sets.

8c. What are MLOps? brief different stages that are involved in the MLOps lifecycles 5M

MLOps stands for Machine Learning Operations. MLOps is focused on streamlining the process of
deploying machine learning models to production, and then maintaining and monitoring them. MLOps is a
collaborative function, often consisting of data scientists, ML engineers, and DevOps engineers. The word
MLOps is a compound of two different fields i.e. machine learning and DevOps from software
engineering.

1. ML Development: This is the basic step that involves creating a complete pipeline beginning from
data processing to model training and evaluation codes.
2. Model Training: Once the setup is ready, the next logical step is to train the model. Here,
continuous training functionality is also needed to adapt to new data or address specific changes.

21
 [Type the document title]

3. Model Evaluation: Performing inference over the trained model and checking the
accuracy/correctness of the output results.
4. Model Deployment: When the proof of concept stage is accomplished, the other part is to deploy
the model according to the industry requirements to face the real-life data.
5. Prediction Serving: After deployment, the model is now ready to serve predictions over the
incoming data.
6. Model Monitoring: Over time, problems such as concept drift can make the results inaccurate hence
continuous monitoring of the model is essential to ensure proper functioning.
7. Data and Model Management: It is a part of the central system that manages the data and models. It
includes maintaining storage, keeping track of different versions, ease of accessibility, security, and
configuration across various cross-functional teams.

Any four stages explanation can be awarded marks

Section V
9a. With a neat diagram explain components of Docker 10M

Components of Docker
1. Docker Client

Docker client uses commands and REST APIs to communicate with the Docker Daemon (Server). When
a client runs any Docker command on the Docker client terminal, the client terminal sends these Docker
commands to the Docker daemon. Docker daemon receives these commands from the Docker client in the
form of command and REST API's request.

Docker Client uses Command Line Interface (CLI) to run the following commands -
 Docker build
 Docker pull
 Docker run
2. Docker Registry
Docker Registry manages and stores the Docker images.There are two types of registries in the Docker
Pubic Registry - Public Registry is also called as Docker hub.
Private Registry - It is used to share images within the enterprise.
3.Docker Daemon: This is the background process that runs on the host machine and manages the
containers. It is responsible for creating, starting, stopping, and removing containers, as well as managing
their network and storage resources.
4.Docker Images
Docker images are the read-only binary templates used to create Docker Containers. It uses a private
container registry to share container images within the enterprise and also uses public container registry to
share container images within the whole world. Metadata is also used by docket images to describe the
container's abilities.

22
 [Type the document title]

4.Docker Containers
Containers are the structural units of Docker, which is used to hold the entire package that is needed to run
the application. The advantage of containers is that it requires very less resources.In other words, we can
say that the image is a template, and the container is a copy of that template.

9b. Demonstrate Simple Linear Regression consideringa dataset that has two variables: salary
(dependent variable) and experience (Independent variable) 10M

To implement the Simple Linear regression model in machine learning using Python, we need to follow the
below steps:
Step1: import libraries
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.linear_model import LinearRegression
from sklearn.model_selection import train_test_split

step2 :Load the dataset

df = pd.read_csv("salary_data.csv")

Step3: pre-processing.Check for any missing values and handle it by any suitable method
df.isnull().sum()
mean_A = df['salary'].mean()
df['salary'] = df['salary'].fillna(mean_A)
df.isnull().sum()

Step4: Split the data set. Extract the dependent and independent variables from the given dataset.
x = df['Experience']
y = df['salary']
x_train, x_test, y_train, y_test = train_test_split(x,y, train_size = 0.8, test_size=0.2, random_state = 21)
x_train=x_train.values.reshape(-1,1)
x_test =x_test.values.reshape(-1,1)

Step5: Build model

model = LinearRegression()
model.fit(x_train, y_train)

Step6: Find the accuracy

model.score(x_test,y_test)
10a.Summerize any two cloud deployment models 10M
1.Public Cloud
The public cloud makes it possible for anybody to access systems and services. The public cloud may be
less secure as it is open to everyone. The public cloud is one in which cloud infrastructure services are
provided over the internet to the general people or major industry groups. The infrastructure in this cloud
model is owned by the entity that delivers the cloud services, not by the consumer. It is a type of cloud
hosting that allows customers and users to easily access systems and services. This form of cloud
computing is an excellent example of cloud hosting, in which service providers supply services to a
variety of customers. In this arrangement, storage backup and retrieval services are given for free, as a
subscription, or on a per-user basis. Example: Google App Engine etc.
Advantages of Public Cloud Model:
 Minimal Investment: Because it is a pay-per-use service, there is no substantial upfront fee, making it
excellent for enterprises that require immediate access to resources.

23
 [Type the document title]

 No setup cost: The entire infrastructure is fully subsidized by the cloud service providers, thus there
is no need to set up any hardware.
 Infrastructure Management is not required: Using the public cloud does not necessitate infrastructure
management.
 No maintenance: The maintenance work is done by the service provider (Not users).
 Dynamic Scalability: To fulfil company’s needs, on-demand resources are accessible.
Disadvantages of Public Cloud Model:
 Less secure: Public cloud is less secure as resources are public so there is no guarantee of high-level
security.
 Low customization: It is accessed by many public so it can’t be customized according to personal
requirements.
2.Private Cloud
The private cloud deployment model is the exact opposite of the public cloud deployment model. It’s a
one-on-one environment for a single user (customer). There is no need to share your hardware with
anyone else. The distinction between private and public clouds is in how you handle all of the hardware.
It is also called the “internal cloud” & it refers to the ability to access systems and services within a
given border or organization. The cloud platform is implemented in a cloud-based secure environment
that is protected by powerful firewalls and under the supervision of an organization’s IT department. The
private cloud gives greater flexibility of control over cloud resources.
Advantages of Private Cloud Model:
 Better Control: You are the sole owner of the property. You gain complete command over service
integration, IT operations, policies, and user behaviour.
 Data Security and Privacy: It’s suitable for storing corporate information to which only authorized
staff have access. By segmenting resources within the same infrastructure, improved access and
security can be achieved.
 Supports Legacy Systems: This approach is designed to work with legacy systems that are unable to
access the public cloud.
 Customization: Unlike a public cloud deployment, a private cloud allows a company to tailor its
solution to meet its specific needs.

Disadvantages of Private Cloud Model:

 Less scalable: Private clouds are scaled within a certain range as there is less number of clients.
 Costly: Private clouds are costlier as they provide personalized facilities.
Hybrid Cloud
By bridging the public and private worlds with a layer of proprietary software, hybrid cloud computing
gives the best of both worlds. With a hybrid solution, you may host the app in a safe environment while
taking advantage of the public cloud’s cost savings. Organizations can move data and applications
between different clouds using a combination of two or more cloud deployment methods, depending on
their needs.
Advantages of Hybrid Cloud Model:
 Flexibility and control: Businesses with more flexibility can design personalized solutions that meet
their particular needs.
 Cost: Because public clouds provide scalability, you’ll only be responsible for paying for the extra
capacity if you require it.
 Security: Because data is properly separated, the chances of data theft by attackers are considerably
reduced.
Disadvantages of Hybrid Cloud Model:
 Difficult to manage: Hybrid clouds are difficult to manage as it is a combination of both public and
private cloud. So, it is complex.
 Slow data transmission: Data transmission in the hybrid cloud takes place through the public cloud so
latency occurs.

24
 [Type the document title]

10b.Discuss any five ethical challenges in AI 10M

1. Bias and Discrimination: AI systems can perpetuate and even amplify biases and discrimination if
they are not properly designed and tested. For example, an AI system that is used to make decisions
about hiring or lending may discriminate against certain groups of people if it is trained on data that
contains such biases.
2. Privacy and Security: AI systems can collect and process large amounts of personal data, which can
raise concerns about privacy and security. For example, an AI system that is used to monitor
people's behaviour or predict their behaviour may collect sensitive information about them, which
could be used to discriminate against them or to cause harm.
3. Lack of Explain ability and Transparency: Many AI systems are based on complex algorithms that
are difficult to understand or explain. This can make it difficult for people to understand how the AI
system is making its decisions and to hold the system accountable for its actions.
4. Job Loss: AI systems can automate many tasks that are currently performed by humans, which can
lead to job loss and other economic dislocation. It's important to consider the social and economic
impacts of AI and to ensure that the benefits of AI are shared fairly.
5. Autonomy and Control: AI systems can be programmed to make decisions and take actions
autonomously, which can raise concerns about who is in control of the system and who is
responsible for its actions.
6. Ethical dilemmas: AI systems may be faced with ethical dilemmas, such as the trade-off between
human lives and property damage in self-driving cars, and it may be difficult for the system to make
the right decision.
7. Societal impact: The development and deployment of AI can have a significant impact on society,
and it's important to consider the broader ethical, social, and political implications of AI and to
ensure that the technology

Certified that model answers prepared by me for code 20CS51I are from syllabus and scheme of
valuation prepared by me is correct

SAVITHA KAMALAPURKAR
Lecturer Dept of Comp Science &Engg
S.J(Govt.) Polytechnic Bangalore

25