Book Recommendation System using Collaborative

Filtering

AD Aakash Nanthan
16BCE1061
Daniel Jebakumar
17BLC1080
VIT CHENNAI

ABSTRACT:

Recommendation systems are used for the purpose of suggesting new items
to a user depending on their past activity. They direct users towards items
which can meet their needs by processing a large database of Information.
Various techniques have been introduced for recommending items such as
content, collaborative and association mining techniques.

This project solves the problem of data scarcity by combining the

collaborative-based filtering and association rule mining to achieve better
performance.The results obtained are demonstrated and the proposed
recommendation algorithms perform better and solve the challenges such as
data scarcity and scalability.

KEYWORDS:

Collaboration, Data mining, Prediction algorithm, Recommender systems,

collaborative filtering, data mining.

INTRODUCTION:
A recommender system is a type of information filtering system that predicts
the rating or preference that a user would give to an item and uses it to
recommend similar items that the user might rate similarly. Recommendation
Systems might be targeted to individuals who do not have enough personal
experience to evaluate the potentially overwhelming alternatives that a web
site might offer. Some companies such as Amazon, Facebook and Google are
already utilizing the power of recommender systems in building their new and
existing systems. As examples, it is common to rely on peer
recommendations when selecting a book to read; employers use
recommendation letters in their recruiting decisions; and when selecting a
movie to watch, individuals read and rely on the movie reviews that a critic
has written. This approach is popularly known as collaborative-ltering and its
aim is to ensure that if the active user agreed in the past with some users,
then the other recommendations coming from these similar users should be
relevant as well and of interest to the active user. In this report, we introduce a
novel architecture for the implementation of a recommender system algorithm
which can be used to improve the recommendation of books to users.

METHODOLOGY:

In this report, we discuss a specific way to implement a recommendation

system using the R programming language. Our model uses Collaborative
Filtering and Associative Rule Mining to recommend new items to the user
depending on the user’s past activity.
The main steps followed in this method are:

1. Analyzing and cleaning the data

2. Selecting a subset of users to work with
3. Explore the data to find areas of interest
4. Visualize the new data
5. Recommend new items (books) to the user

PROCEDURE ADOPTED:

Collaborative filtering is a technique applied mainly in recommendation

systems to make predictions about the interests of a user by gathering
inclination data from multiple users. It works on the simple idea that if User A
shows a similar pattern of interest in certain items to User B, then User B’s
choice of items can also be shown to User A as they have a higher chance of
liking this item. For example, a collaborative filtering recommendation
system for television tastes could influence forecasts about which
television show a client should like given a halfway rundown of that
client's tastes (likes or dislikes), which is what we have adopted.
STEPS TO BE FOLLOWED:

Using ​recommenderlab​: ​recommenderlab​ is an R-package that provides the

infrastructure to evaluate and compare several collaborative-filtering algorithms.
Many algorithms are already implemented in the package, and we can use the
available ones to save some coding effort, or add custom algorithms and use the
infrastructure (e.g. cross validation).

Find similar users:​ For this step we select users that have in common that they
rated the same books.

Get predictions for other books​: In order to get recommendations for our user
we would take the most similar users and average their ratings for books and In
order to get recommendations for our user we would take the most similar users
and average their ratings for books.

We can also improve the algorithm by:

1. Weighting the ratings by similarity​ - This means that the more similar a

user is to the current user the more weight his/her ratings receive in the

calculation of the predictions.

2. Weighting the similarity calculation​ - Depending on the books these

users co-rated. The more books these users co-rated the more reliable is

their similarity score.

RESULTS AND DISCUSSION:

​ ​Fig.1​ ​ ​Fig.2

Fig.3​ ​ ​ ​Fig.4

​Fig.5
Fig.6 Fig.7

Fig.8

The Best Rated Book of all time

OBSERVATIONS:

From Fig.1, it is quite apparent that most books have ratings of 4 and 5,
From Fig.2, we can see that a lot of users have rated only 0 to 10 books,from
Fig.3, we can see that the average rating is around 4, from Fig.4, it is clearly
visible that at least around 20 people rate the book and from Fig.5, it is visible
that the average user rates 4 for any book.

Fig.6, tells us that Fantasy and Romance are the most rated books out of all the
books, fig.7, tells us that a majority of the books are in english and finally Fig.8,
tells us that the longer the title length, the higher the average rating. We also
found that The Hunger Games is the most rated book, followed by Harry Potter.
CONCLUSION:

We identified some interesting trends using this dataset. In summary, the

observed effects on book rating are rather small, suggesting that book ratings
are mainly driven by other aspects, hopefully including the quality of the book
itself. In collaborative filtering the main idea is to use ratings from similar users
to create recommendations. The basic algorithm is easy to implement by hand,
but there are some packages like ​recommenderlab w ​ hich greatly simplify our
work.
REFERENCES:

1. Gomez-Uribe, C.A., Hunt, N.: The Netflix recommender system:

algorithms, business value, and innovation.

2. Davidson, J., Livingston, B., Sampath, D., Liebald, B., Liu, J., Nandy, P.,
Vleet, T.V., Gargi, U., Gupta, S., He, Y., Lambert, M.: The YouTube video
recommendation system.

3. Jacobson, K., Murali, V., Newett, E., Whitman, B., Yon, R.: Music
personalization at Spotify. In: Proceedings of the 10th ACM Conference
on Recommender Systems (RecSys ’16). ACM, New York.

4. Sarwar, B., Karypis, G., Konstan, J., Reidl, J.: Item-based collaborative
filtering recommendation algorithms, Proceedings of the tenth
international conference on World Wide Web - WWW 01 (2001).

5. Linden, G., Smith, B., York, J.: Amazon.com recommendations:

item-to-item collaborative filtering.

6. Bobadilla, J., Ortega, F., Hernando, A., Gutirrez, A.: Recommender

systems survey. Knowl.-Based Syst.

7. Karypis, G.: Evaluation of Item-Based Top-N Recommendation

Algorithms. In Proceedings of the tenth international conference on
Information and knowledge management (CIKM ’01).

8. Hu, Y., Koren, Y., Volinsky, C.: Collaborative Filtering for Implicit
Feedback Datasets.

9. Liu, N.N., Xiang, E.W., Zhao, M., Yang, Q.: Unifying explicit and implicit
feedback for collaborative filtering.