0% found this document useful (0 votes)
1 views17 pages

DSP Module 3 - Part 2

The presentation by Dr. N. Yuvaraj covers linear regression and logistic regression, demonstrating how to predict exam scores based on study hours and whether a student passes an exam based on hours studied. It includes calculations for model parameters, predictions, and interpretations of results, highlighting the importance of accuracy and potential improvements for logistic regression models. The document also features visual representations of data and predictions.

Uploaded by

Divya Meena S
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
1 views17 pages

DSP Module 3 - Part 2

The presentation by Dr. N. Yuvaraj covers linear regression and logistic regression, demonstrating how to predict exam scores based on study hours and whether a student passes an exam based on hours studied. It includes calculations for model parameters, predictions, and interpretations of results, highlighting the importance of accuracy and potential improvements for logistic regression models. The document also features visual representations of data and predictions.

Uploaded by

Divya Meena S
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Presentation by: Dr. N.

Yuvaraj , SCOPE,VIT

Problems on linear regression

PRESENTATION BY: YUVARAJ , SCOPE, VIT


Presentation by: Dr. [Link] , SCOPE,VIT
Linear regression
A teacher wants to predict the exam score of students based on the
number of hours they study using simple linear regression.
Fit a simple linear regression model using the
Hours Exam form: Y=β0+β1X where:
Student
Studied (X) Score (Y) 1. Y = predicted exam score
A 1 35 2. X = hours studied
3. β0, β1 are intercept and slope
B 2 40
1. Calculate the values of β0 and β1 using the least
C 3 50 squares method.
D 4 55 [Link] the model to predict the exam score for:
E 5 65 1. a student who studies 3.5 hours
2. a student who studies 7 hours
F 6 70
[Link]: The regression line & Actual data points
4. Interpret the meaning of β1 in this context.
2

PRESENTATION BY: YUVARAJ , SCOPE, VIT


Presentation by: Dr. [Link] , SCOPE,VIT
Linear regression
Solution: Calculate the values of β0 and β1 using the least squares method.
Hours Exam
Studied (X) Score (Y)
1 35
2 40
3 50
4 55
5 65
6 70

PRESENTATION BY: YUVARAJ , SCOPE, VIT


Presentation by: Dr. [Link] , SCOPE,VIT
Linear regression
Solution: Calculate the values of β0 and β1 using the least squares method

PRESENTATION BY: YUVARAJ , SCOPE, VIT


Presentation by: Dr. [Link] , SCOPE,VIT
Linear regression
Solution: Calculate the values of β0 and β1 using the least squares method

PRESENTATION BY: YUVARAJ , SCOPE, VIT


Presentation by: Dr. [Link] , SCOPE,VIT
Linear regression
2. Use the model to predict the exam score for:
1.a student who studies 3.5 hours
2.a student who studies 7 hours
1. For 3.5 Hours

2. For 7 Hours

3. Interpretation of β1=7.29
For every additional hour a student studies, their predicted exam
score increases by approximately 7.29 points.
6

PRESENTATION BY: YUVARAJ , SCOPE, VIT


Presentation by: Dr. [Link] , SCOPE,VIT
Linear regression

Blue points: Actual student scores


Red line: Predicted regression line
Green & Orange dots:
Predictions for 3.5 and 7 hours
respectively

PRESENTATION BY: YUVARAJ , SCOPE, VIT


Presentation by: Dr. [Link] , SCOPE,VIT

Logistic Regression

PRESENTATION BY: YUVARAJ , SCOPE, VIT


Presentation by: Dr. [Link] , SCOPE,VIT

Logistic Regression – Example 1


¡ Predict if a student passes (1) or fails (0) an exam based on hours studied
using logistic regression.

Hours Result Logistic Regression Model


Student
Studied (X) (Y)
A 1 0
B 2 0
C 3 0
D 4 1
We want to find the best values for β0 and
E 5 1
β1 that maximize the likelihood.
F 6 1
9

PRESENTATION BY: YUVARAJ , SCOPE, VIT


Presentation by: Dr. [Link] , SCOPE,VIT

Logistic Regression – Example 1


¡ Predict if a student passes (1) or fails (0) based on hours studied

Initial try with : β0=−6 ; β1=1.2


X Y z = -6 + p = 1 / (1 +
(Hours) (Actual) 1.2X e^(-z))
1 0 -4.8 0.0082
2 0 -3.6 0.0266
3 0 -2.4 0.0832
4 1 -1.2 0.2315
5 1 0.0 0.5000
6 1 1.2 0.7685
10

PRESENTATION BY: YUVARAJ , SCOPE, VIT


Presentation by: Dr. [Link] , SCOPE,VIT

Logistic Regression – Example 1


¡ Compute Log-Likelihood

X Y p Log-likelihood term
1 0 0.0082 log (1−0.0082)=log (0.9918)=−0.0082
2 0 0.0266 log (0.9734)= -0.0269 logL≈−0.0082−0.0269
−0.0866−1.464−0.693
3 0 0.0832 log (0.9168)= -0.0866 1−0.2633
4 1 0.2315 log (0.2315)= -1.464
LLH =−2.5421
5 1 0.5 log (0.5)=−0.6931
6 1 0.7685 log (0.7685)=−0.2633 11

PRESENTATION BY: YUVARAJ , SCOPE, VIT


Presentation by: Dr. [Link] , SCOPE,VIT

Logistic Regression – Problem 2

12

PRESENTATION BY: YUVARAJ , SCOPE, VIT


Presentation by: Dr. [Link] , SCOPE,VIT

Logistic Regression – Problem 2


Use Case: Predicting Online Purchase Based on Time Spent on Website
A company uses logistic regression to predict whether a visitor will make a purchase (1 =
Yes, 0 = No) based on the time (in minutes) they spend on the website.
Model Parameters: β0=−2.5 ; β1=0.3

Visit Time on Site Actual Calculate the probability of purchase for each visitor
or (X) Purchase using the logistic regression formula.
1 2 No (0)
2 4 No (0)
3 5 Yes (1) Classify each visitor as Yes (1) or No (0) based on
4 6 Yes (1) the threshold: If p≥0.5⇒Yes ; Else⇒No
5 8 Yes (1) Compare the predictions with the actual
6 10 No (0) purchases, and compute:
7 12 Yes (1) •Number of correct predictions 13

•Accuracy (%)
PRESENTATION BY: YUVARAJ , SCOPE, VIT
Presentation by: Dr. [Link] , SCOPE,VIT

Logistic Regression - Problem


Given: Model Parameters: β0=−2.5 ; β1=0.3

Need to calculate:

X (Time p=1/ Prediction


Visitor z = -2.5 + 0.3·X Actual Correct?
on Site) (1 + e^-z) (p≥0.5)
1 2 -1.9 0.130 0 (No) 0 Yes
2 4 -1.3 0.214 0 (No) 0 Yes
3 5 -1.0 0.269 0 (No) 1 No
4 6 -0.7 0.331 0 (No) 1 No
5 8 -0.1 0.475 0 (No) 1 No
6 10 0.5 0.622 1 (Yes) 0 No 14

7 12 1.1 0.750 1 (Yes) 1 Yes


PRESENTATION BY: YUVARAJ , SCOPE, VIT
Presentation by: Dr. [Link] , SCOPE,VIT

Logistic Regression - Problem


Correct Predictions: 3 out of 7 (Visitors 1, 2, 7)
Incorrect Predictions: 4 out of 7
Accuracy: 3/7×100≈42.86%

Inferences
•The model correctly identifies short time visitors as non-buyers.
•But it fails to predict correctly for medium-range buyers (5–8 mins).
•One long-time visitor (10 mins) is wrongly predicted as a buyer.
This suggests the model may need:
•More data
•More features (like age, pages viewed, device type)
•Or parameter tuning to improve prediction performance 15

PRESENTATION BY: YUVARAJ , SCOPE, VIT


Presentation by: Dr. [Link] , SCOPE,VIT

Logistic Regression - Problem

•Blue curve: The logistic function


showing the predicted probability of
purchase for any time value.
•Dashed gray line at p = 0.5:
Classification threshold above this is
predicted as "Yes", below is "No".
•Dots on the curve:
• Green: Correct prediction
• Red: Wrong prediction
• Circle (o): Actual purchase (1)
• Cross (x): No purchase (0)

16

PRESENTATION BY: YUVARAJ , SCOPE, VIT


Presentation by: Dr. [Link] , SCOPE,VIT

Thank You

17

PRESENTATION BY: YUVARAJ , SCOPE, VIT

You might also like