Scribd
Upload
43 views

Book Recommendation System-Capstone Project 4

RECOMMENDATION

Uploaded by

saritbarua29
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
43 views

Book Recommendation System-Capstone Project 4

RECOMMENDATION

Uploaded by

saritbarua29
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
You are on page 1/ 31

Capstone Project - III

Team 3 :
BOOK RECOMMENDATION SYSTEM

Team Members

Syed Sharin
Shyam Sundar K
Fathima K
Content

● Problem statement
● Data Summary
● Analysis of different datasets
● Data Cleaning
● Outlier treatment
● Imputing missing values
● Different Recommendation Model
● Challenges
● Conclusion
● Future Scope
Problem Statement

During the last few decades, with the rise of


Youtube, Amazon, Netflix, and many other such
web services, recommender systems have
become much more important in our lives in
terms of providing highly personalized and
relevant content.

The main objective is to create a


recommendation system to recommend
relevant books to users based on popularity
and user interests.
Data Summary
The dataset is comprised of three csv files:: User_df, Books_df, Ratings_df

Users_dataset.
● User-ID (unique for each user)
● Location (contains city, state and country separated by commas)
● Age Shape of Dataset - (278858, 3)

Books_dataset.
● ISBN (unique for each book)
● Image-URL-S
● Book-Title
● Image-URL-M
● Book-Author
● Image-URL-L
● Year-Of-Publication
● Publisher ● Shape of Dataset - (271360, 8)

Ratings_dataset.
● User-ID ● Book-Rating
● ISBN ● Shape of Dataset - (1149780, 3)
Observations from Users_df (Age)

● The Age range given here is from 0


To 250.
● Outliers in the Age column.
Observations from Users_df (Age)

● The Age range distribution is


right skewed
● Most active readers lie in age
group 20- 40
Observations from Users_df (Location)
● Splitting Location column and analysing country.
● Most active readers are from USA.
Observations from Book_df (Authors)
Agatha Christie wrote highest number of books in our given dataset
Observations from Book_df (Publishers)
Harlequin published highest number of books in our given dataset
Observations from Ratings_df (Book_Rating)

● Higher ratings are more common amongst users


● Rating 8 has been rated the highest number of times
Data Cleaning
1. Null Value Imputation:

Age column has 40% missing values


Imputing missing values
● Outliers in Age column
● Age has positive Skewness (right tail) so we can use median to fill Nan values,
Data Cleaning
1. Null Value Imputation:
Replacing strings by int values
Different Models
1.)Popularity Based Recommendation

Book weighted average formula:

Weighted Rating(WR)=[vR/(v+m)]+[mC/(v+m)]

Where,

v is the number of votes for the books;


m is the minimum votes required to be listed in the chart;
R is the average rating of the book; and
C is the mean vote across the whole report.
Different Models
Different Models
2.)Model based collaborative filtering

SVD NMF
Different Models
SVD Model Results
Different Models
SVD Model Results
Different Models
SVD Model Results
Different Models
User-ID - 193458
Test set: predicted top rated books
Different Models

Test set: actual top rated books


Collaborative Filtering-(Item-Item based)
3.)Collaborative Filtering-(Item-Item based)

● Cosine Similarity
● Nearest Neighbour
Different Models
SVD and Correlation
Recommendations for Harry Potter and the Sorcerer's Stone (Book 1)

Input
Output
Different Models
4.)Collaborative Filtering-(User-Item based)
Different Models
Model Results
Conclusion
● In EDA, the Top-10 most rated books were essentially novels. Books like The Lovely Bone
and The Secret Life of Bees were very well perceived.

● Majority of the readers were of the age bracket 20-35 and most of them came from North
American and European countries namely USA, Canada, UK, Germany and Spain.

● If we look at the ratings distribution, most of the books have high ratings with maximum
books being rated 8. Ratings below 5 are few in number.

● Author with the most books was Agatha Christie, William Shakespeare and Stephen King.

● For modelling, it was observed that for model based collaborative filtering SVD technique
worked way better than NMF with lower Mean Absolute Error (MAE) .
Conclusion
• A recommendation system helps an organization to create loyal customers.
The recommendation system today are very powerful that they can handle the
new customer too who has visited the site for the first time. They recommend
the products which are currently trending or highly rated and they can also
recommend the products which bring maximum profitto the company.
A book recommendation system is a type of recommendation system where we have to
recommend similar type of books to the reader based on his interest. The books
recommendation system is used by online websites which provide ebooks like google
playbooks, open library, good Read’s, etc.
Challenges

● Handling of sparsity was a major challenge as well since the user interactions were
not present for the majority of the books.

● Understanding the metric for evaluation was a challenge as well.

● Since the data consisted of text data, data cleaning was a major challenge in
features like Location etc..

● Decision making on missing value imputations and outlier treatment was quite
challenging as well.
Future Scope

● Given more information regarding the books dataset, namely features like Genre,
Description etc, we could implement a content-filtering based recommendation
system and compare the results with the existing collaborative-filtering based
system.

● We would like to explore various clustering approaches for clustering the users
based on Age, Location etc., and then implement voting algorithms to recommend
items to the user depending on the cluster into which it belongs.
Thank You

You might also like