A Final Year Project

Presentation
On
“MOVIE RECOMMENDATION SYSTEM
USING ITEM BASED COLLABORATIVE FILTERING”

Anuj Pokharel (Roll No. 11082/073)

Dhiraj Kafle (Roll No. 11087/073)
Dipendra Lal Shrestha (Roll No. 11090/073)
Resuka Manandhar (Roll No. 11103/073)
1. Introduction
2. Problem Definition
Contents
3. Objective
4. Research Methodology
4.1 Literature Review
4.2 Types of Collaborative Filtering
4.3 Algorithm
4.4 System Flowchart
5. System Development & Design
5.1 System Architecture
5.2 System Requirements
5.3 System Design
6. Data Collection
7. Testing
8. Result & Discussion
9. Limitation
10. Conclusion
2
 1. Introduction

• Recommendation system are information filtering system which

generally uses collaborative filtering algorithms .

• Collaborative filtering is a technique of making predictions (filtering)

about the interests of a user by collecting preferences such as ratings.

• Collaborative filtering is dependent on correlation rate between users

or items.

3
Cont.…

4
 2. Problem Definition

• Lots of movies with different genres are widely available.

• Not having the appropriate means of recommending the movies.

• Recommendation system gives relief on that sector of the movies sites

for the users.

5
 3. Objective

• To create movie rating system where user can rate a movie.

• To develop recommendation system that suggest user specific

interesting movies based on the rating given by the user.

6
 4. Research Methodology

4.1. Literature Review

• Below are the literature review paper titles:

Movies Recommendation System Using Collaborative Filtering

and K-means.

Movie Recommendation System Using Collaborative Filtering.

A Survey on Recommendation System with Collaborative Filtering

Using Big Data.

7
4.2 Types of Collaborative Filtering

Collaborative filtering can be categorized into two main methods.

They are:
i. User-based Collaborative Filtering
ii. Item-based Collaborative Filtering

8
 Cont…
Item-based Collaborative Filtering
Items
  I1 I2 … In-1 In
Items

I1 R1,1 R1,2 … R1,n-1 R1,n

I2 R2,1 R2,2 … R2,n-1 R2,n

… … … … … …
Im-1 Rm-1,1 Rm-1,2 … Rm-1,n-1 Rm-1,n

Im Rm,1 Rm,2 … Rm,n-1 Rm,n

Table 1: Item Similarity Matrix

9
4.3 Algorithm

• The Pearson correlation coefficient has a high statistical significance.

• The correlation coefficient formula finds out the relation between the variables.
• It returns the values between -1 and 1.

Where:
xi = x values
yi = y values
x̅ = Mean of x values
y̅ = Mean of y values
(xi – x̅ )2 = Deviation Squared
(yi – y̅ )2 = Deviation Squared
10
 4.4 System Flowchart

Figure 1: Flowchart of Movie Recommendation System 11

 5. System Development & Design
5.1. System Requirements

Figure 2: Use Case Diagram

12
 5.2. System Design

Figure 3: Activity Diagram 13

 6. Data Collection

• Primarily, the rating datasets are from ‘Movie Lens’, which is

officially provided from grouplens.org site.

• Datasets have 1,00,000 real world user’s ratings to 9,000 movies.

• Secondarily, the static data for displaying purpose such as movie

image, title, genre, imdb rating obtained by performing web scraping
of the IMDb website.

14
 7. Testing
• Three testings: Unit, Integrated, Compatibility
 Unit testing: testing of individual units
For eg:
Objective Sign Up for new user
Input Username, E-mail Address,
Password, Re-Password
Expected Output Sign Up & able to access the
system
Error • Please fill the field
• The two password fields didn’t
match.
• A user with that username
already exists.

Table 2: Sign Up Test 15

 Cont.…
Integrated Testing:

• Group of components combined to produce output.

• Components: Sign Up, Login, Add Movies, Delete User, Rate

Movie, Search Movie, Filter Movies, General Recommendations,
Recommendation

 Compatibility Testing:
• By running the project into different browsers.

• Even in live with cloud web hosting server with domain:

https://trinityproject.khullasikshya.com
16
8. Result & Discussion

Figure 4: Correlation Coefficient Line Graph 17

 Cont....
Chosen Movie User Top 6 Recommended Correlation Execution
Rating Movies Coefficient Time

    Inception 0.5413  
 
     
Django Unchained 0.5353
Interstellar 5  
  Edge of Tomorrow 0.5296 0.5 second
 
The Dark Knight Rises 0.5283

The Imitation Game 0.5236

Deadpool 0.5077

Table 3: Recommended Movies List For Interstellar 18

 9. Limitations

• It does not perform sentiment analysis which would give more

accurate results.

• It doesn’t provide user to user-based recommendation.

• With the addition of new movie, there occurs cold start problem for
recommendation to those newly added latest movies.

19
 10. Conclusion

• Recommendation of the movies on the basis of ratings.

• Item- Item based collaborative filtering approach along with Pearson

Correlation Coefficient algorithm.

• System recommends similar movies.

20
Thank You !

21