Vous êtes sur la page 1sur 6

 

 
Advertisement Click-Through-Rate 
Optimisation using reinforcement 
learning 
 
 
Data Analytics
Semester 7
BE 1

BY:
Amey Kolhe 4144
Mandar Pande 4156

 
 
 

1
Contents

Problem Statement 3

Abstract 3

Objective 3

Outcome 3

Introduction 4

Project Development Outline 4

Scope 5

Test Cases 5

Results 5

Conclusion 6

REFERENCES 6

2
Problem Statement

Click-through rate (CTR) is the ratio of users who click on a specific link to the number of
total users who view a page, email, or advertisement. It is commonly used to measure the
success of an online advertising campaign for a particular website as well as the
effectiveness of email campaigns. Given a set of advertisements, determine which one will
have the highest CTR and provide optimal results using reinforcement learning.

Abstract

In online advertising, the click-through rate (CTR) is the percentage of individuals viewing a
web page who click on a specific advertisement that appears on the page. Click-through rate
measures how successful an ad has been in capturing users' interest. The higher the
click-through rate, the more successful the ad has been in generating interest. A high
click-through rate can help a website owner support the site through advertising dollars.
In this project we use the dataset provided at superdatascience.com to train our model. We
employ reinforcement learning algorithm namely, Upper Confidence Bound algorithm to
optimise the click through rates of the ads.

Objective

The mini project aims to attain the following objectives:

● Clean data by using various data cleaning techniques


● Pre-process the data so that it can be used for analysis and statistical computation
● Analyse provided data present data in a readable format for any user
● Use UCB model to optimise the CTR of the ads.

Outcome

After completing this mini-project, we will be able to build Upper Confidence Bound by
performing data preprocessing and testing for accuracy on the test data.

3
Introduction

A click-through rate can help digital marketers measure the efficacy of a variety of online marketing
campaigns. It may be utilized with a variety of mediums, such as display advertisements, email
advertising and paid search.

It can also be used to measure the effectiveness of advertising copy, titles and descriptions that
make up the metadata of online content. Since most websites are built to compel users to take an
action, click-through rates can help digital marketers learn what works and what doesn't.

Such monitoring does not give advertisers any insight into the intent and reasoning of an individual
who has clicked on an online ad or piece of content but remains a standard because of its
convenience.

Because of its simplicity and efficiency, optimising CTR becomes an important task. In this project,
we will use reinforcement learning algorithm to optimise the click through rates of an ad company

Project Development Outline

For ​the successful implementation of the project we undertook the following activities:

1. Brainstorming for project field selection


2. Requirement gathering for the selected topic
3. Requirement analysis of the collected requirements
4. Finalizing requirements and problem statement.
5. Brainstorming for data analysis techniques
6. Incorporation of changes
7. Implementation of reinforcement learning model
8. Testing of model
9. Final testing
10. Result Analysis
11. Project Report Submission

4
Scope

Dataset from the superdatascience.com has various ad versions on horizontal and users on
vertical. Each time a user clicks an advertisement, the value of that cell is 1, else 0. We only
explore the upper confidence bound algorithm in this project as it provides satisfactory
results. The result is analyzed using a histogram. The UCB algorithm has been implemented
by us for the given dataset.

​Test Cases

Case ID Case Expected Actual Result


Output Output

1 Compile UCB Model Compilation Compilation Pass


Successful Successful

2 Prediction using UCB Model Prediction Prediction Pass


complete complete

3 Generate Histogram Histogram Histogram Pass


generated generated
Successful Successful

Results

Fig 1 : Histogram of ads selections

5
Conclusion

We have successfully completed the project with the required functionalities and stated
scope. Results have been demonstrated using histogram for better comparison and analysis.

REFERENCES

[1] ​https://scikit-learn.org/​ (Date visited : 13/9/19 time : 15:30)


[2] ​https://matplotlib.org/​ (Date visited : 15/9/19 time : 15:30)
[3] ​https://en.wikipedia.org/wiki/Data_analysis​ (Date visited : 11/9/19 time : 17:30)

Vous aimerez peut-être aussi