Module Code & Module Title

CU6051NT Artificial
Intelligence

Assessment Weightage & Type

25% Individual Coursework

Year and Semester

2021-22 Autumn

Student Name: Shailesh Bhattarai

London Met ID:20048503
College ID: np05cp4s210130@iic.edu.np
Assignment Due Date: 14 December 2022
Assignment Submission Date:
Title: Product Recommendation System
Word Count (Where required):

I confirm that I understand my coursework needs to be submitted online via Google Classroom under
the relevant module page before the deadline in order for my assignment to be accepted and marked.
I am fully aware that late submissions will be treated as non-submission and a mark of zero will be
awarded.
Abstract

Because of the rapid expansion of the internet and social media in recent years.
Because of its ability to provide improved products, a product recommendation is vital
in our social lives. Users can be recommended a set of products based on their
interests or the popularity of the products. Even though several product
recommendation systems have been presented, most of them either cannot
recommend a product to existing users efficiently or cannot recommend a product to a
new user at all. We offer a product recommendation system in this research that can
recommend products to both new and existing users.

Recommender systems, also known as recommendation systems, are information

filtering systems that are typically connected with a variety of consumer and
commercial applications. These systems operate as a link between numerous content
providers, such as social media websites, e-commerce portals, streaming platforms,
and app users, by proposing items from the app database that match the user's tastes
and previous activity. Such personalized systems play a vital role, especially when the
user is unclear of the item to the searched for.
CS6P05NT ARTIFICAL INTELLIGENCE

1. Introduction

1.1. Explanation of the topic/AI concept used

Artificial intelligence

Building intelligent computers that can carry out tasks that traditionally require human intelligence
is the goal of artificial intelligence (AI), a broad field of computer science. Machine intelligence is
referred to as artificial intelligence. Artificial intelligence has risen to prominence in the modern
world. Computers that have been taught to recognize and imitate human behavior are simulating
human intelligence. These machines are capable of making mistakes and growing from them to
perform things that humans do. As artificial intelligence (AI) advances, it will significantly affect
our quality of life. Everyone today wants to interact with AI technology in some capacity, whether
as a consumer or as a professional working in the industry.

The phrase is frequently used to describe a project that aims to develop systems with
intellectual capabilities similar to those of humans, including the capacity for reasoning,
decoding meaning, generalizing, and learning from prior experiences. It has been
demonstrated that computers may be programmed to easily accomplish exceedingly
hard tasks, such as discovering proofs for mathematical theorems or playing chess,
since the introduction of the digital computer in the 1940s. Although computer
processing speed and memory capacity continue to rise, no program has yet been able
to match human adaptability in broader sectors or in tasks requiring a significant
amount of common knowledge. [ CITATION WEr20 \l 1033 ]

1
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

Machine learning

Artificial intelligence uses machine learning to imitate human intelligence. The

computer builds a propensity model utilizing algorithms and previous data in order to
understand how to react to particular behaviors. Propensity models will then start to
make predictions after that (like scoring leads or something). These are some of the
most often used marketing applications and functions, however AI is capable of much
more. And even though it might seem like machines are about to take over, most of the
work still needs to be done by people. By include people in email automation and
letting AI take care of much of the hard lifting while we concentrate on other tasks, we
use AI primarily to save time. [ CITATION KKe18 \l 1033 ]

2
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

AI concept used in recommendation system

Recommendation systems have been built using artificial intelligence (AI), notably
computer intelligence and machine learning methodologies and algorithms, to increase
prediction accuracy and address data sparsity and cold start issues. It extensively
examines a variety of AI-related recommender system-related challenges and
examines how these systems have been improved by applying techniques including
fuzzy techniques, transfer learning, genetic algorithms, evolutionary algorithms, neural
networks and deep learning, and active learning. [ CITATION Rog19 \l 1033 ]

Offering goods and services that specifically address the demands of a specific
customer is difficult for businesses in a competitive market. By helping to address a
significant issue—information overload—personalized e-services improve user
experience and make it easier for customers to make decisions. The recommender
systems used in these customized e-services were developed twenty years ago
employing methods and theories from several artificial intelligence (AI) disciplines for
locating user profiles and preferences.

Recently, several AI techniques have been used to recommender systems, which has improved
user pleasure and experience. A higher standard of suggestion is possible using AI than with
conventional techniques. A new age for recommender systems has begun as a result of this,
enabling deeper insights into user-item relationships, the presentation of more complex data
representations, and the discovery of thorough information in demographical, textual, virtual, and
contextual data.

3
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

1.2 Explanation/introduction of the chosen problem domain/topic

The recommendation engine can be used to increase earnings, sales, and revenues in
general. It is a great marketing tool, especially for e-commerce. With eleven additional
studies demonstrating the benefit of recommendation engines in the e-commerce
industry, it is no wonder that customized product recommendations are so well-liked in
the retail industry. A recommendation engine is a system for filtering information that
uploads data in accordance with users' choices, interests, or past behavior related to
an item. Based on their profile, it can predict a user's interest in a given item. Thanks to
product suggestion algorithms, customers may quickly and simply find what they're
looking for. Few recommendation systems have been created thus far to find items that
the user has already viewed, purchased, or in some other manner interacted with. A
recommendation system must be flexible enough to change as user behavior does. It
must to be able to function in a dynamic environment, giving users timely notice of
promotional offers, assortment adjustments, and price changes. [ CITATION FOI15 \l
1033 ]

Algorithms used in recommendation systems typically rely on past purchases and page
visits. As a result of using artificial intelligence to analyze user interactions and select
aesthetically appealing products that will appeal to any specific customer, many firms
now provide real-time suggestions. Artificial intelligence is used by recommendation
engines to provide quick, precise recommendations that are tailored to the needs and
preferences of each consumer.

By offering suggestions based on a user's aesthetic tastes rather than product

descriptions, artificial intelligence is assisting in the improvement of internet searches.
By guiding consumers in finding products or material that they might not otherwise be
able to find, artificial intelligence consulting engines seem to be posing a challenge to
search fields. As a result, modern websites like Amazon, Facebook, YouTube, and
others depend heavily on recommendation algorithms.[ CITATION bap19 \l 1033 ]

4
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

2. Background

2.1 Research work done on the chosen topic/problem domain

A product recommender system is a piece of technology that suggests which products

should be displayed to users of a brand's digital assets using machine learning. As a
result of a number of algorithmic findings, recommendation algorithms mine user,
product, and contextual data both online and offline to offer each user a customized
experience. Customers can find what they are seeking for and, in some circumstances,
items they aren't even aware they are looking for by improving the discovery process.
Businesses may gain a better understanding of each user's unique preferences and
interests in this way, enabling them to enhance real-time performance while also
enhancing their long-term testing roadmaps. There is no one-size-fits-all strategy for
product recommendations that marketers should use for every device. Different
approaches must be employed for various individuals depending on the quantity of
information that is available about the client, their activity, and the context of the
products on a website. It includes details on the site's activity, status, geographic
location, the time of day, previous transactions, and more. Additionally,
recommendations can offer critical insights and the chance to better understand a
consumer in order to better understand customers, add value, and enhance the entire
relationship with a business. [ CITATION Lea20 \l 1033 ]

The recommender system filters the most crucial information from a large quantity of data using
the user's data as well as additional factors like the user's preferences and interests. In order to
provide recommendations, it evaluates the user and object's compatibility as well as their shared
characteristics.

5
20048503 Shailesh
Bhattarai
 CS6P05NT ARTIFICAL INTELLIGENCE

Purpose of the Recommendation System

i. Finding products that interest them is advantageous for users.

ii. help product suppliers distribute their goods to the right customers

iii. Products that are most pertinent to users can be found by users.

iv. content that is personalized for each user

v. Assist websites in increasing user engagement.

There are two methods to build a recommendation system:

i) Content Based Recommendation system

ii) Collaborative Filtering Recommendation system

In my project I will be using Collaborative Filtering System.

6
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

2.1.1 Collaborative Filtering System

Collaborative Filtering, the most well-known application suggestion engine, is based on

educated assumptions that customers who have enjoyed a product will do so again.
Another name for this kind of algorithm is a collaborative shift with a focus on products.
Instead of objects in this filtering, users are screened and linked to each user. In this
approach, only user behavior is taken into account. Relying just on their profile and
content information is insufficient. The actions of Users who rate a product favorably
are associated with the actions of other Users who offer the same rating. [ CITATION
Man17 \l 1033 ]

Collaborative filtering requires a list of items depending on the user's prior selections.
This system only requires a few product characteristics to work. Each object and user
have a unique embedding or feature vector that describes them both and sinks them in
the same embedding location. It creates enclosures for users and goods on its own.
Based on the users' locations, collaborative filtering suggests well-known, interesting,
or well-liked content. As the population grows more diversified, the published tales will
reflect a greater level of public interest. Another illustration of collaborative filtering in
action is Wikipedia. Collaborative filtering does not require content extraction or
analysis. People will be able to properly assess information when compared to
counting functions. Multimedia or complex objects, such music, movies, and images,
start to work properly. [ CITATION Tua21 \l 1033 ]

7
20048503 Shailesh
Bhattarai
 CS6P05NT ARTIFICAL INTELLIGENCE

Figure 1: collaborative filtering syst

2.1.1 Example of Collaborative Filtering System

One of the best applications of collaborative filtering is in the area of e-commerce.

You'll notice that an e-commerce website will provide you suggestions for other things
when you visit. There are some items there that are almost exact replicas of what
you're looking for. Now, you might be wondering how the website identifies your
interests. Collaborative filtering is to blame for everything.

The use of friends' suggestions is also very common on social networking sites. A
highly helpful feature on Facebook, for instance, is the People you may know section,
which presents a list of people who can be joined as friends. According to this method,
if you are friends with the 10th out of the 11th densely associated persons, you must
be friends with the 11th. This method educates and calculates missing edges based on
social connection data. Social connections are made via collaborative filtering
algorithms.

Let's examine another situation. Amshu and Sailesh both like to play the same games.
Nikki like playing it and using it. Amshu hasn't played that game yet, but the algorithm
has determined that Amshu and Sailesh have a similar interest, so it recommends that
Sailesh play it. Recommender systems are capable of collaborative filtering using the
same product. The other buyer will also find the item appealing. [ CITATION MDE11 \l

8
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

1033 ]

9
20048503 Shailesh
Bhattarai
 CS6P05NT ARTIFICAL INTELLIGENCE

Figure 2: Example of collaborative filtering system

2.1.1. Types of collaborative filtering

There are two types of collaborative filtering. They are:

Memory based collaborative filtering

Memory-based methods make use of the data you already have (likes, votes, clicks, etc.) to create
correlations (similarities) between individuals (Collaborative Filtering) or items (Content-Based
Recommendation) in order to suggest an item, I to a user u who hasn't seen it before. In the case
of collaborative filtering, we obtain recommendations from items seen by users nearby you, hence
the phrase collaborative. The goal of content-based recommendation, on the other hand, is to offer
similar new products by comparing items based on their features (e.g., movie genre, actors, book
publisher or author, etc.). Simple similarity measurements (Cosine similarity, Pearson correlation,
Jaccard coefficient, etc.) are frequently used in memory-based systems to group similar individuals
or objects together. Memory-based algorithms use similarity measurements on two vectors (rows
or columns) of such a matrix to obtain a number reflecting similarity when there are users on one
dimension and things on the other, with the cells carrying votes or likes.
There are two types of Memory based algorithm. They are:

10
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

 Item-Item based algorithm

 User-User based algorithm

 Model based collaborative filtering

On the other hand, model-based approaches try to complete the gaps in this matrix.
They take on the issue of "predicting" a user's level of interest in something they
haven't before seen. They achieve this by training a model that can predict the user's
rating for a new item that has just been added to the system using a number of
machine learning techniques on a user's vector of objects. In this approach, models
are built utilizing various data mining and machine learning algorithms to predict user
ratings of unrated items. There are numerous model-based collaborative filtering
methods available. The Bayesian network, the Baye Latent semantic models,
clustering models, and singular value decomposition Latent Dirichlet Models of
Allocation, Markov decision-making, and multiple multiplication factor are examples of
probabilistic latent semantic analysis.

There are two types of Models based algorithm. They are:

 Matrix factorization based

 Clustering based

11
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

Figure 3: types of collaborative filtering

2.1. Analysis of existing work

Computer programs called recommender systems provide recommendations to users

based on a number of factors. These systems predict the product that consumers are
most likely to purchase and be interested in. Recommender systems are used by
Netflix, Amazon, and other businesses to help their customers locate the ideal book or
movie.

Collaborative filtering is a crucial machine learning technique that aids a computer in

content selection based on previous interactions and user-provided data. Using the
user's prior data, collaborative filtering algorithms provide comparable results. Based
on the user's historical data, collaborative filtering algorithms offer comparable results..[
CITATION XSu09 \l 1033 ]

The role of collaborative filtering recommendations system at various real-world can

be seen from the example of :

12
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

1. Facebook

13
20048503 Shailesh
Bhattarai
 CS6P05NT ARTIFICAL INTELLIGENCE

Social networking, which aims to connect individuals from all over the world, was invented by
Facebook, a social networking site, in 2004. The social networking site Facebook, which is
presently run by Mark Zuckerberg, uses a number of AI techniques to improve it. On the other side,
one of the most amazing tactics this social media juggernaut does is collaborative filtering.
"Facebook uses a variety of recommender systems for different sections of the website. To give
information it believes will be helpful to the user, one algorithm, for instance, might be utilized in the
user timeline while other recommender systems are used in the News and Marketplace areas.
Collaborative Filtering is therefore utilized by Facebook.

2. Amazon
When Amazon was first established in 1994, it only sold books. Amazon incorporates a
variety of the leading machine learning techniques, including collaborative filtering, into its
recommendation engine for greater performance and increased user engagement. To further
provide high-quality recommendations, our platform chooses item-based collaborative filtering
over a user-based strategy. Item-based collaborative filtering on the platform has generally
proven to be a successful strategy that has boosted the company's prospects for financial
success. Collaborative filtering initially only used one method, a user-based method.
However, Amazon was the one to create an item-based strategy that started to focus on
things rather than users.

3. Solution

3.1. Explanation of the proposed solution/approach to

solving the problem

The Recommendation may be very helpful in a number of industries, including e-

commerce, movies, and other ones, as was previously said. The technology that is
created will be advantageous to these sectors of the economy. Customers will be able to
receive movie recommendations from technology. The experiment will be based on a
mock dataset, and an algorithm will be used to process the data and construct the tests.
I'll be employing this structure. I'll be employing a collaborative filtering recommendation

14
20048503 Shailesh
Bhattarai
 CS6P05NT ARTIFICAL INTELLIGENCE

system in this framework. Any good or service has a stickiness element that makes
people want to buy it. Its proprietary algorithms are made to foresee a user's interest,
recommend various items to the user in a variety of ways, and maintain that interest
throughout.

Recommendation systems help users in several organizations, including Numerous online

merchants use recommender systems to increase sales using machine learning (ML). By
adopting and implementing this strategy on their websites, many retail businesses produce a
significant number of sales. Netflix and Amazon, two forerunners in the use of recommenders,
have released their algorithms for recommendation systems to entice customers. All of these
organizations have put the recommendations system into place to make it easier for customers
to do business with them and to boost marketing efforts. [ CITATION ite15 \l 1033 ]

15
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

3.2. Algorithm

A set of instructions called an algorithm tells a computer how to turn a collection of

world facts into knowledge that is useful. When describing an activity for which a set of
rules has been created, the word algorithm is usually used in conjunction with other
descriptive words. A method for assessing what kind of information can be recovered
from a large amount of data is a search algorithm, for instance. A collection of
guidelines called an encryption algorithm is used to encrypt data or messages so that
only authorized individuals may decipher them. [ CITATION AED20 \l 1033 ]

Memory-based collaborative filtering can be done in two different ways: item-to-

item and user-to-user. For this project, I'd like to employ an item-item based
collaborative filtering method. The User-User Based Memory Collaborative
Filtering Approach was rejected for a number of reasons, including the market's
extremely low user rating rate, particularly when a new product is made available
on the internet. will debut with a poor rating and no straightforward option for
customers to contact the product. If we cooperate, we may do an analysis of the
current market situation. Instead of using user-based pricing, many organizations
favor item-based pricing. Therefore, I decided on item-Item based memory
collaborative filtering for my project.

Item-Item based algorithm

It was created and initially used by Amazon in 1998. Instead of connecting the user to
similar consumers, item-to-item collaborative filtering links every item the user has
purchased and rated to related items before combining those related products into a
recommendation list. The initial stage in creating the model is to identify similarities
between each pair of items. There are various methods for determining how similar two
item pairs are. Utilizing cosine similarity is one of the most widely used techniques.

A model-based approach for recommendations is item-based collaborative filtering. Using one of

several techniques, the computer figures out the correlations between various components in the

16
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

dataset. Using a variety of similarity tests, these similarity values are then used to estimate the
user-item pair scores. By examining every object, the degree of similarity is calculated. The
ratings that people who rated both products gave each one determines how comparable the two
items are. [ CITATION Sau21 \l 1033 ]

17
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

Figure 4: Item-Item based algorithm

18
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

Following is the process of working mechanism of item-item based algorithm.

Item to Item Similarity

19
20048503 Shailesh
Bhattarai
 CS6P05NT ARTIFICAL INTELLIGENCE

The first stage in creating the model is determining which item pairs are
comparable to one another. There are various methods for determining how
similar two item pairs are. Utilizing cosine similarity is one of the most widely
used techniques.

Formula for Cosine Similarity:

Figure 5: formula of cosine similarity

Prediction Computation

Executing a recommendation system is the second stage. It generates ratings using the items
(already evaluated by the user) that are most comparable to the missing item. We therefore
attempt to produce predictions based on the reviews of comparable products. We calculate this
using a technique that uses the weighted sum of the ratings of other comparable products to
determine the rating for a specific item.

Figure 6: formula of prediction computation

Example of item-item based algorithm using collaborative filtering.

A set table with various things and the users who have rated such items is provided below. The
evaluation is straightforward and ranges from 1 to 5. The rating given by an ith User to a jth Item
is indicated by each entry in the table. Since most users only rate a small number of goods, most
cells are typically vacant. In this case, 3 things and 4 users are seized. We need to track out the
user's missing ratings.

20
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

User/Item Item_1 Item_2 Item_3

User_1 2 - 3
User_2 5 2 -
User_3 3 3 1
User_4 - 2 2

Table 1: Example of item-item based algorithm

Step 1: Finding similarities of all the item pairs.

(Item 1, Item 2), (Item 1, Item 3), and (Item 2, Item 3) are the item pairings. picking out

each item to pair one at a time. Then, we locate every user who has given ratings for

both items in the item pair. Using the aforementioned cosine calculation, determine

how similar the two things are.

calculating the similarity between the two items using the cosine formula as stated

above. Sim (Item1, Item2)

In the table, we can see only User_1 and User_2 have rated for both items 1 and 2.

Thus, let I1 be vector for Item_1 and I2 be for Item_2. Then,

I1 = 5U2 + 3U3 and,

I2 = 2U2 + 3U3

Figure 7: Value of s (I1, I2)

So, the value of s (I1, I2) is 0.9

Sim (Item2, Item3)

21
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

In the table we can see only User_3 and User_4 have rated for both the items 1 and 2.

22
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

Thus, let I2 be vector for Item_2 and I3 be for Item_3. Then,

I2 = 3U3 + 2U4 and,

I3 = 1U3 + 2U4

Figure 8: Value of s (I2, I3)

So, the value of s (I2, I3) is 0.869

Sim (Item1, Item3)

In the table we can see only User_1 and User_3 have rated for both the items 1 and 2.

Thus, let I1 be vector for Item_1 and I3 be for Item_3. Then,

I1 = 2U1 + 3U3 and,

I3 = 3U1 + 1U3

Figure 9: Value of s (I1, I3)

So, the value of s (I1, I3) is 0.869

Step 2: Generating the missing ratings in the table

Now, in this step we calculate the ratings that are missing items in the

table.

Rating of Item_2 for User_1

23
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

Figure 10: rating of User 1 Item 2

Rating of Item_3 for User_2

Figure 11: rating of User 2 Item 3

Rating of Item_1 for User_4

Figure 12: rating of User 4 Item 1

Therefore, the ratings that are missing in the table of U1I2 is 2.49, U2I3 is 3.43 and
U4I1 is 2.0

3.3. Pseudocode
24
20048503 Shailesh
Bhattarai
 CS6P05NT ARTIFICAL INTELLIGENCE

One approach to creating or planning a program is pseudocode. Pseudocode is false code since
pseudo means false. The words phony or imitation would be a better translation. False
Pseudocode (not the real thing). Although it mimics or appears like genuine code, it is not. It
explains the goals of a program using English statements. Because there is no compiler that can
convert pseudocode to any machine language, it is fictitious. Pseudocode is used to record the
design of a program or module (also known as the algorithm). (Busbee, 2018)

The team of designers and programmers can review pseudocode to make sure that actual
programming is likely to adhere to design standards because it is thorough yet comprehensible.
Finding faults early in the development process is less expensive than doing it later. The
acceptable pseudocode is then recast in a programming language's vocabulary and grammar.
Sometimes approaches based on computer-aided software engineering are combined with
pseudocode. (Admin, 2005)

25
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

START

IMPORT System

IMPORT NumPy

IMPORT library

READ dataset

PROCESS dataset

TAKES the dataset (80% of training data,20%of test data)

TRAIN the data

CREATE list of product Ids

DEFINE User Ratings

VISUALIZE most popular product

DEFINE Similarity

EXTRACT Similarity

USE the algorithm of collaborative filtering

TRAIN the above algorithm

PREDICT the data

CHECK the Missing Values

EXECUTE

26
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

27
20048503 Shailesh
Bhattarai
 CS6P05NT ARTIFICAL INTELLIGENCE

3.4. Flowchart

A flowchart is a diagram that shows the actions, choices, and sequences of a workflow or process.
A basic flowchart is the most basic type of a process map, even though there are many distinct
forms of flowcharts. It is a potent tool that may be applied to many different fields for process
planning, visualization, documentation, and improvement. (Asana, 2022)

Collaborative Filtering Systems:

Uses information about a set of users and their relations with the item to provide
recommendations to the active user.

28
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

29
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

30
20048503 Shailesh
Bhattarai
 CS6P05NT ARTIFICAL INTELLIGENCE

The collaborative filtering can be adapted with neighborhood

31
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

methods, whose focus is on relationship between the items or,

alternatively, between the users. They are:

I. User-based
The collaborative filtering can be adapted with neighborhood
methods, whose focus is on relationship between the items or,
alternatively, between the users. They are:

I. User-based

32
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

User-based

A flowchart is a diagram that depicts an algorithm. It is frequently used by

programmers as a problem-solving technique. It employs symbols that are linked
together to represent the flow of information and processing. Flowcharting is the
process of creating a flowchart for an algorithm. A graphical representation of an
information system's or program's operation sequence. Flowcharts for information
systems depict how data moves from source documents to the computer and then to
users. The sequence of instructions in a single program or function is shown in a
program flowchart. Each sort of flowchart has its own set of symbols.[ CITATION
gee202 \l 1033
]

33
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

Figure 13: flowchart of product recommendation system

34
20048503 Shailesh
Bhattarai
CS6P05NT ARTIFICAL INTELLIGENCE

4. Conclusion

4.1. Analysis of Work Done

In conclusion, collaborative filtering is more than just a way to generate suggestions.

How the internet sees us and serves us to the best of its ability is what matters most.
Some people could think it's intrusive. However, it is brilliant for a machine to learn our
purchasing habits, make recommendations depending on who is using them, and
make choices that closely resemble our own.

My topic of choice is the Product Recommendation System, a popular AI idea in the

industry. a straightforward prototype for guiding customers toward recommended
products or related items to those they have looked for. for instance, the consumer will
be shown other brands or options after conducting a mobile phone search. The
keyword filtering suggests recommending low-cost cell phones. The context of this
research is the recommendation system.

The algorithm used to produce the project had a variety of benefits and drawbacks.
Due to collaborative filtering, the content of the items, such as the type or genre of the
film, may not always tell the entire story. We can predict the item rating even in the
absence of any knowledge about the item and without if the consumer will purchase it.
Collaborative filtering is the application where the recommendation algorithm shines.

35
20048503 Shailesh
Bhattarai
 CS6P05NT ARTIFICAL INTELLIGENCE

4.2. The Solution to Real-World Problems

As internet users, we all interact with product suggestion algorithms practically every day through
Google searches, music or movie streaming services, online purchasing, social media reading, and
dating apps.

Product recommendation systems have become one of the most popular and effective machine
learning applications in the business world as a result. Recent research from Monette shows that
product recommendations can increase purchase rates by 70%, both in the initial session and in
following sessions, as well as average order values by 33%. According to a Salesforce study,
customers who click on product recommendations had 4.5x higher basket rates, 4.8x more product
views per visit, and a 5x higher per-visit expenditure.

Platforms for e-commerce use collaborative filtering. A consumer might get product
recommendations that are catered to their preferences when exploring products on an e-commerce
website like Amazon, for instance. This can help buyers choose the right product and make
informed decisions about it. Additionally, e-commerce companies profit from the suggestion system
because it allows them to consistently boost their sales by a sizeable amount. When a consumer
receives a recommended item, they are likely to choose and purchase it, providing a way to boost
the profit of an unbalanced business.

36
20048503 Shailesh
Bhattarai
 CS6P05NT ARTIFICAL INTELLIGENCE

4.3. Future Work

As this course work is based on sentiment analysis of product reviews, we cover

artificial intelligence (AI) and its significance in recommendation systems in this report.
A study was done to identify the issue, the kinds of technology that should be used, the
algorithms that should be employed, the flowchart and pseudocode for the product
recommendation system, and the findings were studied and provided above. The
introduction, problem, research, and other information are supplied in this document
because part one of this course work is mostly about documentation.

The next step is to put the research findings into practice after it has been thoroughly researched
and documented. To do this, a dataset must first be imported into the application. The project's
research has already been finished, and work on the development phase has begun.

37
20048503 Shailesh
Bhattarai