A Mini Project Report

on

Movie Recommendation System

Submitted to the
Pune Institute of Computer Technology, Pune
In partial fulfillment for the award of the Degree of
Bachelor of Engineering
in
Information Technology
by

Soham Patil 33259

Soham Phatak 33261
Aashlesh Wawge 33283

Under the guidance of

Prof. S. A. Jakhete

Department Of Information Technology

Pune Institute of Computer Technology College of Engineering
Sr. No 27, Pune-Satara Road, Dhankawadi, Pune - 411 043.

2023-2024

i
 SCTR’s PUNE INSTITUTE OF COMPUTER TECHNOLOGY
DEPARTMENT OF INFORMATION TECHNOLOGY

CERTIFICATE

This is to certify that the project report entitled

MOVIE RECOMMENDATION SYSTEM

Submitted by

Soham Patil 33259

Soham Phatak 33261
Aashlesh Wawge 33283

is a bonafide work carried out by them under the supervision of Prof. S. A. Jakhete and it is approved
for the partial fulfillment of the requirement of Laboratory Practice I for the award of the Degree of
Bachelor of Engineering (Information Technology)

Prof. S. A. Jakhete Dr. A. S. Ghotkar Dr. S. T. Gandhe

Project Guide HOD, IT Principal
PICT, Pune PICT, Pune PICT, Pune

Place:
Date:
ii
 ACKNOWLEDGEMENT

We thank everyone who has helped and provided valuable suggestions for successfully developing
a wonderful project.
We are very grateful to our guide Prof. S. A. Jakhete, Head of Department Dr. A. S. Ghotkar and
our Principal Dr. S. T. Gandhe. They have been very supportive and have ensured that all facilities
remained available for smooth progress of the project.
We would like to thank our professor and Prof. S. A. Jakhete for providing very valuable and
timely suggestions and help.

iii
 ABSTRACT

This project implements a movie recommendation system using a dataset containing

information on different movies, including movie IDs, titles, and details about their cast
and crew. The system leverages machine learning techniques such as content-based
filtering, which uses movie metadata like cast and crew information to recommend movies
similar to those the user has previously watched. The goal is to enhance user engagement
by providing personalized movie suggestions based on specific features extracted from the
dataset. By processing and analyzing large amounts of movie-related data, the
recommendation system offers tailored recommendations that align with users' viewing
preferences.

iv
 CONTENTS

1. INTRODUCTION
1.1 Purpose, Problem statement 01
1.2 Scope, Objective 01

2. LITERATURE SURVEY
2.1 Introduction 02
2.2 Detail Literature survey 02
2.3 Findings of Literature survey 02

3. SYSTEM ARCHITECTURE AND DESIGN

3.1 Detail Architecture 03
3.2 Dataset Description 04
3.3 Detail Phases 04
3.4 Algorithms 05

4. EXPERIMENTATION AND RESULTS

4.1 Phase-wise results 06
4.2 Explanation with example 10
4.3 Accuracy 10
4.4 Tools used 10

5. CONCLUSION AND FUTURE SCOPE

5.1 Conclusion 11
5.2 Future scope 11

6. REFERENCES

7. ANNEXURE
7.1 Implementation Code 00

v
 LIST OF TABLES

Table Number Table Title Page Number

3.1 Dataset Description 04

LIST OF FIGURES

Figure Number Table Title Page Number

3.1 System Architecture of Movie Recommendation System 03
4.1 Data Preprocessing 06-07
4.2 Feature Engineering 07-08
4.3 Model Training 09
4.4 Recommendation Generation 09

vi
 1. INTRODUCTION
1.1 PROBLEM STATEMENT
The purpose of this project is to develop a movie recommendation system that addresses the
challenge of content discovery in an increasingly vast and diverse film landscape. By utilizing movie
metadata such as cast, crew, and genres, the system aims to provide personalized suggestions
tailored to individual user preferences. The goal is to enhance user engagement and satisfaction by
ensuring that viewers can easily find movies that align with their interests, ultimately improving
their overall viewing experience. This project also seeks to explore the effectiveness of content-
based filtering techniques in generating relevant recommendations.

1.2 SCOPE, OBJECTIVE

▪ To develop a system that offers personalized movie suggestions based on user preferences,
enhancing the viewing experience.
▪ To leverage various attributes of movies, including cast, crew, genres, and descriptions, in order
to create a comprehensive recommendation engine.
▪ To implement and evaluate content-based filtering techniques that analyze movie features,
allowing users to discover films similar to those they have previously enjoyed.
▪ To incorporate user interaction and feedback mechanisms to continually improve the
recommendation accuracy and relevance of suggestions.

1
 2. LITERATURE SURVEY
2.1 INTRODUCTION

The literature survey provides a comprehensive overview of existing research and methodologies
related to movie recommendation systems. This section highlights the evolution of recommendation
technologies, focusing on key approaches such as collaborative filtering and content-based filtering.
By analyzing prior studies, we aim to identify trends, challenges, and advancements in the field,
establishing a foundational understanding for the development of an effective movie
recommendation system.

2.2 DETAILED LITERATURE SURVEY

Recent studies have demonstrated the effectiveness of collaborative filtering techniques in

enhancing the accuracy of movie recommendations. Collaborative filtering leverages user behavior
data, enabling the system to identify patterns among users with similar tastes. By analyzing
interactions, such as ratings and viewing history, these models can predict user preferences, thereby
offering tailored suggestions. However, while effective, collaborative filtering methods often face
challenges such as the cold start problem, where new users or items lack sufficient data for accurate
predictions. This limitation necessitates the exploration of complementary techniques to enhance
recommendation quality.

In contrast, content-based filtering approaches focus on the intrinsic features of the movies
themselves. By analyzing attributes such as genre, cast, and plot descriptions, content-based systems
can recommend films that align closely with the user's past preferences. This method is particularly
advantageous for users with well-defined tastes, as it provides more consistent recommendations
based on specific characteristics. Nonetheless, content-based filtering may struggle with novelty, as
it often leads to suggestions that are too similar to previously enjoyed items, potentially limiting
user engagement over time.

2.3 FINDINGS OF LITERATURE SURVEY

The literature survey reveals a growing trend towards hybrid recommendation systems that combine
both collaborative and content-based filtering methods to leverage the strengths of each approach.
By integrating diverse data sources, these hybrid models aim to enhance recommendation accuracy
and overcome common challenges such as the cold start problem and limited novelty. Overall, the
findings underscore the importance of continuous research and development in the field, as well as
the potential for innovative solutions that can improve user experience in movie recommendation
systems.

2
 3. SYSTEM ARCHITECTURE AND DESIGN
3.1 DETAIL ARCHITECTURE

Fig. 3.1 System Architecture of Movie Recommendation System

▪ Data Ingestion: This involves collecting and loading data from the specified dataset, which in
this case is the TMDB movies dataset. The data will be cleaned and preprocessed to ensure
quality and consistency.

▪ Feature Engineering: Extract relevant features from the dataset, such as budget, genres,
revenue, and vote average. This step may include encoding categorical variables and
normalizing numerical features.

▪ Model Selection: Choose suitable machine learning algorithms based on the problem type (e.g.,
regression or classification). This could involve algorithms like Linear Regression, Decision
Trees, or Random Forests.

▪ Training Phase: Split the dataset into training and testing sets. Train the selected models using
the training set while tuning hyperparameters for optimal performance.

▪ Deployment: Once a satisfactory model is achieved, deploy it for predictions on new data. This
could involve creating a web service or integrating it into an application.

3
3.2 DATASET DESCRIPTION

The dataset used in this project is the TMDB 5000 Movies dataset and the TMDB 5000 Credit
datset. These two datasets are combined together for the model training. Key characteristics include:

TMDB 5000 Movies Dataset

COLUMN NAME DESCRIPTION

Budget The financial budget allocated for the movie, represented in dollars.
Genres A list of genres associated with the movie, often represented as JSON objects
containing genre IDs and names.
Homepage The official website of the movie, providing additional information and
promotional content.
Id A unique identifier for each movie in the dataset.
Keywords A list of keywords related to the movie, represented as JSON objects that
describe themes or elements.
Original_language The language in which the movie was originally produced.
Original_title The title of the movie as it was originally released.
Overview A brief summary or description of the movie's plot and themes.
Popularity A numerical value indicating the popularity of the movie, often based on user
interactions and ratings.
Production_companies A list of companies involved in producing the movie, often represented as
JSON objects with company names and IDs
Production_countries The countries where the movie was produced, represented as JSON objects
with country names and ISO codes.
Release_date The date when the movie was officially released, formatted as YYYY-MM-
DD.
Revenue The total revenue generated by the movie, represented in dollars.
Runtime The total duration of the movie in minutes.
Spoken_languages A list of languages spoken in the movie, represented as JSON objects
containing language codes and names.
Status The current status of the movie (e.g., Released, Post Production).
Tagline A catchy phrase or slogan associated with the movie, often used for marketing
purposes.
Title The title of the movie as it appears to audiences.
Vote_average The average rating given to the movie by viewers, typically on a scale from 1
to 10.
Vote_count The total number of votes or ratings submitted by viewers for the movie.

Table 3.1 Dataset Description

3.3 DETAILED PHASES

The project can be divided into several distinct phases:

▪ Data Collection: Import the dataset and examine its structure.

▪ Data Preprocessing:

▪ Handle missing values.

4
 ▪ Normalize numerical features and encode categorical variables.

▪ Exploratory Data Analysis (EDA):

▪ Visualize data distributions and relationships between features.

▪ Identify trends and patterns that may inform feature selection.

▪ Model Development:

▪ Select appropriate algorithms based on the analysis.

▪ Train models using cross-validation techniques.

▪ Model Evaluation:

▪ Use metrics to evaluate model performance on test data.

▪ Fine-tune models based on evaluation results.

▪ Deployment: Prepare the model for real-world application through a suitable interface.

3.4 ALGORITHMS
For this project, the following algorithms can be implemented:

▪ Linear Regression: For predicting continuous outcomes such as revenue based on budget and
other features.

▪ Decision Trees: Useful for classification tasks such as predicting movie genres based on various
attributes.

▪ Random Forests: An ensemble method that improves prediction accuracy by combining

multiple decision trees.

▪ Support Vector Machines (SVM): Effective for classification problems where a clear margin
of separation exists between classes.

By employing these algorithms, one can effectively analyze the TMDB dataset and derive
meaningful insights or predictions regarding movie performances.

5
 4. EXPERIMENTATION AND RESULTS
4.1 PHASE-WISE RESULTS
The experimentation phase of the project can be divided into several key stages, each yielding
specific results:

▪ Data Preprocessing: The dataset was cleaned and transformed, resulting in a structured format
ready for analysis. Missing values were addressed, and categorical variables were encoded.

6
 Fig. 4.1 Data Preprocessing
▪ Feature Engineering: Relevant features were extracted, enhancing the dataset's usability for
the recommendation system. This included parsing genres and keywords into usable formats.

7
 Fig. 4.2 Feature Engineering

▪ Model Training: Various algorithms were tested, including content-based filtering and
collaborative filtering, leading to the identification of the most effective model

8
 Fig. 4.3 Model Training
▪ Recommendation Generation: The model successfully generated movie recommendations
based on user-inputted keywords, demonstrating its capability to match user preferences with
available movies.

Fig. 4.4 Recommendation Generation

9
4.2 EXPLAINATION WITH EXAMPLE
For instance, when a user inputs the keyword "space adventure," the model retrieves movies that
include this theme in their genres or keywords. The recommendation system may suggest titles like
"Avatar" and "John Carter," which feature elements of space exploration and adventure. This
illustrates how the model effectively aligns user interests with relevant movie attributes.

4.3 ACCURACY
We can't directly calculate accuracy in the traditional sense for a recommendation system like this.
Accuracy in recommendation systems is typically measured by metrics like:
▪ Precision: How many recommended items were relevant to the user.
▪ Recall: How many relevant items were recommended out of all relevant items.
▪ NDCG (Normalized Discounted Cumulative Gain): Considers the order of recommendations
and their relevance.
▪ MAP (Mean Average Precision): Measures the average precision across multiple users or
queries.
▪ Hit Rate: Whether a relevant item was present within the top-k recommendations.

To assess the model's accuracy, we would need a dataset with user ratings or preferences for movies.
We could then:
1. Split the data into training and testing sets.
2. Train the recommendation model on the training data.
3. Generate recommendations for users in the test set.
4. Compare the recommended movies with the actual movies that the users liked or rated highly.
5. Calculate the aforementioned metrics to evaluate the model's performance.

4.4 TOOLS USED

• Programming Language: Python for data manipulation and model development.

• Libraries: Pandas for data processing, Scikit-learn for implementing machine learning
algorithms, and NumPy for numerical operations.

• Visualization Tools: Matplotlib and Seaborn for visualizing data distributions and model
performance.

• Development Environment: Jupyter Notebook for interactive coding and analysis.

These tools collectively facilitated the development, testing, and evaluation of the movie
recommendation system.

10
 5. CONCLUSION AND FUTURE SCOPE
5.1 CONCLUSION

The project effectively implements a movie recommendation system that utilizes the given dataset
to suggest films based on user-provided keywords. By analyzing various attributes such as genres,
keywords, and popularity, the model can identify and recommend movies that align with the user's
interests. This approach not only enhances user experience by providing personalized suggestions
but also leverages data-driven insights to uncover hidden gems within the dataset that users may not
have considered otherwise.

5.2 FUTURE SCOPE

In terms of future scope, the project can be expanded by integrating more advanced natural language
processing techniques to better understand and interpret user queries. Additionally, incorporating
user feedback mechanisms would allow for continuous improvement of recommendations based on
user preferences. Fute iterations could also explore collaborative filtering methods to enhance
recommendations by considering similar users' tastes, thereby creating a more robust and dynamic
movie recommendation system that evolves with changing viewer preferences and trends in the film
industry.

11
 6. REFERENCES

[1] Python documentation for the syntax of ast (Abstract Syntax Trees)
https://docs.python.org/3/library/ast.html

[2] Sci-kit Learn Documentation of CountVectorizer

https://scikit-
learn.org/1.5/modules/generated/sklearn.feature_extraction.text.CountVectorizer.ht
ml

[3] TMDB 5000 Movie Dataset

https://www.kaggle.com/datasets/tmdb/tmdb-movie-metadata

[4] Movie Recommender System Project by CampusX on YouTube

https://youtu.be/1xtrIEwY_zY?feature=shared

12