MULTIREC: THE THREE-IN-ONE RECOMMENDER

AISHWARYA JENA, ARYAN NAMBIAR, ISHAN BHISE, SIDDHARTH KUMAR, SOHAM DEO, YASHI SHRIVASTAVA AND PROF. VIJAY GAIKE

Ajeenkya DY Patil University, Charoli Bk. via Lohegaon, District Pune - 412105, Maharashtra, India
Email: [email protected], [email protected]
Contact: +91-8010195878, +91-961991881
─────────────────────────────────────────────────────────────────────────
Abstract: We have been assigned to do a small project based on our ideas and knowledge. Our team decided to
build a multimedia recommendation system. We aimed to build a movie recommendation system, but we added two
more media recommendation systems, they are music and game. Young people nowadays have access to video
games, movies, and music. We knew platforms that already have recommendation systems and can play the media.
However, our website recommends which type of movie the user can watch, which music they can listen to or the
game they can play. Designing the website needs the design concept for the website, the understanding of the front-
end which mainly includes the design of the webpage. Understanding the back end is a vital part as the entire
recommendation system would work with the backend code. After researching thoroughly, we found that we can
build the front-end using CSS and HTML-5, and the back end can be coded using python. After finding that the
same layout could be built for more than one type of media, our team decided to include music and games. The
database, which contains information about the movies, music, and games, was discovered on the internet. There are
certain limitations to the project. We cannot make a recommendation system for every type of media. Artificial
intelligence will not be used to its full extent in our recommendation system. The recommendation system will be
using the statistics and the information provided in the database to recommend the user of the media they search
recommendations for. This recommendation system will help us in understanding how the database is linked to the
program that has been created; it will also help in understanding the basics of coding. This building of knowledge
will help us in understanding how this recommendation system or another type of programme can be improved in
the future.

Index terms: Recommendation System, Statistics, Database, Python

─────────────────────────────────────────────────────────────────────────

I. INTRODUCTION
would give details of the nearby ambulances
Our team had been given the opportunity to design a available, the number of beds available in each
project that could be useful to our society in some hospital, medical stores that were open at the hour
way. We as a team brainstormed a few ideas as to when the user is checking, etc. This idea seems to be
what we could build or do that would help society. equally excellent, but the issue of this idea was the
feasibility. The scale of collaboration with hospitals,
From our team, the first idea, from our teammate ambulances, medical stores would need to be very
Aryan Nambiar, was to build a movie recommender, vast. At this point as a student, the capability of
the same technology that many media platforms use, building such an app is possible, but many factors
such as Spotify, Netflix, Amazon Prime Video, would not remain in sync and would make it
Pandora, etc. These organisations improve their impossible to finish. Hence, we all continued with the
chances of the user to use their product which is the movie recommender.
app or their website that uses a code that is the
foundation to the recommender which is linked to the The movie recommender will be a stepping stone for
frontends of the apps as well as the website. We all of the team members in understanding how
thought of building just the recommender, which can innovation, planning, implementation, and final
help the user to search for a certain genre or a certain execution plays a great role in making a good project
movie and the details of the movie will be displayed successful.
to the user along with the movies that match the
genre, the main character of the movie, etc. The other
idea that had risen in our group was from Aishwarya
Jena. She had thought of an idea where we would
build a system that would help people when needing
medical assistance. Her idea was to make an app that
AIM They are characteristic information and user-item
interaction. Characteristic information is for the
Our aim of the ED project is to build a website that keywords that have been entered, while user-item
serves the purpose of giving options of entertainment interaction is based on the number of views, ratings,
media based on their liking. This would be done by number of purchases, etc. Then there are also
making recommender for multiple entertainment different types of algorithms used in the
media. recommender. Among them are collaborative,
content-based, and hybrid systems. The content-
based system uses the characteristic information,
POTENTIAL PRODUCT meaning it hypothesizes what the user likes, and
displays it to the user. Collaborative filtering uses the
We decided to build the website with only one data of the interaction of the user with the
recommender, the movie recommender. The backend recommender. The collaborative interaction analyses
consists of the linked database of the movies as well which kind of music the user tends to listen to; based
as the code for the recommendation system. The on that the recommender starts to give suggestions. In
frontend will consist of HTML5 and CSS which will collaborative interaction, there are two ways the
manage and be designed as per our expectations. This recommender analyses the data, first is memory and
potential product we thought of will help users in the second is model. The memory-based
getting the possible movies that they would prefer to recommendation suggests another type of media that
watch, or in the same genre, actor, etc. This product is similar to the media that the user likes. On the
will be free to all and can be accessed by anyone other hand, the model-based suggestions can predict
from the website link that we will share. To expect the user’s top 5 items that they might like the most
the entire recommender to work without any issue, and based on that they recommend new items. The
we will need a domain that will have all strong hybrid system uses both content-based and
connections and does not cause any sorts of issues collaborative filtering to give possibly the results that
like crashing the website or the user is unable to the user might find interesting.
access the recommender.
Python: It is a programming language that can be
used for the development of the web, data science,
FINALISED PRODUCT creating new software prototypes, etc. It can be run
on multiple operating systems like Windows, Linux,
The product we finalised is the same as the movie Mac. It is a programming language that can be used
recommender we had planned on building. However, for multiple purposes. They are used by data analysts,
we planned to extend or diversify the media engineers, scientists, children, etc. It is simpler than
recommender types. We finalised to build a JavaScript and C#.
recommender system that will help the users to find
their possible liking for the media such as movies, HTML-5: It is also known as hypertext markup
music and games. These three media recommenders language revision 5, it is used to build the structure
will be built on the website, we will make new web and presentation of the world wide web contents.
pages for the website so that each of the media This revision is compatible with HTML and
recommenders has its own web page. This will help XHTML- style syntax and a few new features in its
the user to distinguish between each of the markup, new API (application programming
recommenders, that is the music, the game and the interface), XHTML and error handling. It increases
movie recommender. the functionality of the markup language (HTML),
the style (CSS) and better interactivity as well.
BACKGROUND INFORMATION
CSS: It is also known as cascading style sheets. This
Recommendation Systems: It is a strong software helps in describing how the elements present on the
that works based on user’s preferences. This software displayed screen, newspaper, or other media should
recommends products (material, media, etc.) to the be. It helps in controlling multiple web pages at one
user that might be of the user’s liking. They are used time. It helps in designing and placing the elements to
for increasing the chances for sale or increasing the produce a more aesthetic looking webpage, which
usage of the recommender. They make the process of can attract the attention of the users.
searching easier for the user to check the content the
user is interested in. There are two methods by which
the recommendation system checks the information.
PROPOSED DESIGNS AND IDEAS

REQUIREMENT SPECIFICATIONS
Input requirements

Hardware requirements ● Name of all the different recommending

systems on the top of the website
● Personal computer ● Information of each recommending system
● Internet to keep the website working on each of the webpage allocated
● Storage space to compile and store the ● Images/ Gifs
programme data ● Database of the recommending system
● Minimum RAM space 8 Gb for reduced ● Front end design using HTML-5 and CSS
time for compiling codes and programming ● Linking of the backend (recommending
● Processor: Intel i3 or better for faster system) and the front end (website layout)
processing
● Minimum internet speed: 10 Mbps Output requirements

● Sections of the website include:

○ Multirec
Software requirements ○ Bingehub
○ Musichub
● Web browser (e.g.: google chrome) ○ Gamehub
● Python ● Consistent page layout with theme-based
● Jupyter notebook colour and images
● HTML-5 ● Images and text should be displayed in
● CSS various web browsers
● Draw.io ● Image (JPEG), video (GIF)
● Analysing and giving the results matching
System Interaction the data put into the search section by the
user
● The website and recommendation system
should be able to run properly on any Processing
browser that is available, such as Google
Chrome, Firefox, or Internet Explorer. ● Clicking on the headings would lead the
● The website should be able to function with users to another page that would be
the software that was used to having different details and would
programme and design it. With the help of be based on the recommendation system that
the ISP, under the condition that the website matches the heading.
is taken down for maintenance, the internet ● The user may get to the top of the screen by
connection should not break down. clicking the button in the bottom right
● Webspace- to store the database in which all corner.
the details of all the different media ● The recommender will analyse the details
are stored. put in by the user for the media that they are
● Smooth working of hyperlinks searching for, based on that the
● Consistency in the layout of the website recommender would display the results.

● Best possible viewing option- Google

Chrome Security
● Images/ GIFs on the website should load
properly ● Images will not be watermarked as they are
● The recommender should be working stock images and gifs and can be used
without any issues faced without any legal notices

● Will not need any other means of security as

the website free for all
 consuming as well. Very few bugs can be found in
website code. The design can be made unique
compared to many other websites or apps. This
Performance requirements website will be an efficient way of recommending the
movie, music or game based on user’s searches.
Compared to rest solutions, this is also one of the cost
To measure the effectiveness of the website: and time-efficient as well.
● Provide an easily navigable system to WEBPAGES
display the details of the recommenders and
the website. Multirec - In this page of the website there will be
● Create a website with a consistent style details of the recommenders that we have built, then
based on the colours chosen. there will be three different links for each of these
● Create a website with the function of the recommenders. Then there is a widget that will allow
users to go and search all different types of the user to go to the top of the page.
the media recommenders
● All tabs, hyperlinks, web page layout and
Bingehub - This page of the website consists of the
input boxes will be tested on different web
details of what the bingehub as a recommender does.
browsers and platforms.
The page will have the recommender running on it
● All buttons should send the user to the
when the user tries to use the recommender. The
respective webpages
recommender will operate using the backend code as
● The go-to top button on the bottom right of
well as the database of movies we've watched.
every webpage should function properly and
take the user to the top of the page
Musichub - This page of the website consists of the
● The recommenders should be able to display
details of what Musichub is and how it works. This
to the user all the various types of movies,
page will have the recommender for music enabled
music or games based on the type they have
with the help of the spotify database that we are
chosen.
using, this database updates as time passes and so the
genre and coding will suggest new songs based on
JUSTIFICATION OF CHOSEN PRODUCT
the user's liking.
The Final solution discussed together as the team was
Gamehub - This page of the website has the
to build a website. It is a very efficient method for
recommender system that suggests the user games as
publicity of the type of recommenders that our team
per what the user searches. This recommender has a
has built. The website will stay online all the time.
more complex system than the Bingehub and the
This makes it more efficient for our team as well as
Musichub. This recommender will be using the
the users of the website.
collaborative as well as content-based filtering,
basically the hybrid filtering system.
After having a proper discussion of the benefits and
drawbacks of using a website, this option seemed the
SNAPSHOTS
most preferable and appropriate. The benefit of it
over an app would be that the users would not need to
download the recommenders, it is a hassle as it fills
up a certain amount of memory within the device
they are using. They are better than offline
recommenders because we as programmers can
update the list of movies anytime we wish to. Making
an app would be a bigger issue for us as we would
have to integrate many languages just to create the
front end of the app.

Therefore, the website is the best possible given

solution based on the team’s suggestions. The
website can be viewed by everyone, the website does
not need much effort as an app or offline
recommender. Editing the website is less time
 1) “Learn Python Programming.” Programiz,
www.programiz.com/python-programming.
2) Techopedia. “What Is HTML5? - Definition
from Techopedia.” Techopedia.com,
Techopedia, 11 Jan. 2017,
3) www.techopedia.com/definition/1891/html5
. Tryolabs. “Introduction to Recommender
Systems.” Tryolabs Blog, Tryolabs, 9 May
2018, tryolabs.com/blog/introduction-to-
recommender-systems/.
4) Underwood, Corinna. “Use Cases of
Recommendation Systems in Business -
Current Applications and Methods.” Emerj,
Emerj, 4 Mar. 2020,

CONCLUSION

Recommender systems are a powerful new

technology that allows businesses to extract
additional value from their user datasets. These
systems assist consumers in locating products they
wish to purchase from a business. Users gain from
recommender systems because they can locate
products they enjoy. They, on the other hand, benefit
the company by increasing sales. Recommender
systems are quickly becoming a critical element in
online E-commerce. The large number of user data in
existing corporate databases puts recommender
systems under strain, and the rising volume of user
data available on the Web will put them even more
under strain. New technologies are required to
enhance the scalability of recommender systems
substantially.

REFERENCES