Lab Assignment - 6

Assignment-6 week 2 Aug

Learning Objective: Learn Naive Bayes Classifier

1.Problem Statement

A university wants to classify whether an email received by a student is Spam or Not Spam based on the presence or absence of certain words.

Consider the following training data, where:

• 1 = word is present
• 0 = word is absent

Email	Free	Win	Offer	Spam
E1	1	1	1	1
E2	1	1	0	1
E3	0	1	1	1
E4	0	1	0	1
E5	1	0	1	0
E6	1	0	0	0
E7	0	0	1	0
E8	0	0	0	0

Using the above training data, implement a Bernoulli Naive Bayes classifier to classify a new email having the following features:

Free	Win	Offer
1	1	0

Tasks

1. Calculate the prior probabilities of Spam and Not Spam.
2. Calculate the conditional probabilities of each feature given the two classes.
3. Use the Naive Bayes assumption to calculate the probability that the new email is Spam and Not Spam.
4. Predict the class of the new email.
5. Implement the same classification using BernoulliNB from Scikit-learn.
6. Compare the manually calculated prediction with the Scikit-learn prediction.

2.Problem Statement

A college wants to predict whether a student is Eligible or Not Eligible for a scholarship based on categorical attributes.

The following data is collected from 12 students.

Student	Academic Performance	Attendance	Family Income	Eligible
S1	High	Good	Low	Yes
S2	High	Good	High	Yes
S3	Medium	Good	Low	Yes
S4	High	Average	Low	Yes
S5	Medium	Good	High	Yes
S6	Low	Poor	Low	No
S7	Low	Average	High	No
S8	Medium	Poor	High	No
S9	Low	Poor	High	No
S10	Medium	Average	Low	Yes
S11	Low	Average	Low	No
S12	High	Poor	High	No

Using the above dataset, implement a Categorical Naive Bayes classifier to predict whether a new student is eligible for the scholarship.

The new student's details are:

Academic Performance	Attendance	Family Income
Medium	Good	Low

Tasks

1. Calculate the prior probabilities of Eligible = Yes and Eligible = No.
2. Calculate the conditional probabilities of each categorical feature for the two classes.
3. Use the Naive Bayes assumption to calculate the posterior probability for both classes.
4. Predict whether the new student is Eligible or Not Eligible.
5. Implement the same classification using CategoricalNB from Scikit-learn.
6. Compare the manually calculated prediction with the Scikit-learn prediction.
7. Repeat the prediction for the following students and observe the results:

Student	Academic Performance	Attendance	Family Income
A	High	Good	High
B	Low	Poor	High
C	Medium	Average	Low

3.Problem Statement

A college wants to automatically classify student feedback into one of three categories:

• Academic
• Infrastructure
• Placement

The following training data contains the word counts in each feedback message.

Feedback	Exam	Course	Faculty	Lab	Computer	Placement	Job	Interview	Category
F1	2	2	1	0	0	0	0	0	Academic
F2	1	2	2	0	0	0	0	0	Academic
F3	2	1	1	0	0	0	0	0	Academic
F4	0	0	0	3	2	0	0	0	Infrastructure
F5	0	0	0	2	3	0	0	0	Infrastructure
F6	0	0	0	1	2	0	0	0	Infrastructure
F7	0	0	0	0	0	3	2	1	Placement
F8	0	0	0	0	0	2	3	1	Placement
F9	0	0	0	0	0	1	2	2	Placement

Tasks

1. Calculate the prior probability of each category.
2. Calculate the word probabilities for each category using Laplace smoothing.
3. Using the Multinomial Naive Bayes formula, classify the following new feedback represented by its word counts:

Exam	Course	Faculty	Lab	Computer	Placement	Job	Interview
1	1	1	0	0	0	0	0

4. Calculate the posterior probability for each of the three categories.
5. Predict the category having the highest posterior probability.
6. Implement the same classification using MultinomialNB from Scikit-learn.
7. Compare the manually calculated result with the Scikit-learn result.

4.Problem Statement

Implement a Naïve Bayes classifier to categorize text documents into topics using the 20 Newsgroups dataset. Compare the performance of Multinomial Naïve Bayes with Bernoulli Naïve Bayes.

Tasks:

● Load and preprocess the 20 Newsgroups dataset.

● Implement Multinomial Naïve Bayes and Bernoulli Naïve Bayes classifiers.

● Evaluate and compare the performance of both models using metrics such as

accuracy and F1-score.

● Discuss the strengths and weaknesses of each Naïve Bayes variant for text

classification.