Professional Documents
Culture Documents
Movie Recommendation System(全文标明引文)
Movie Recommendation System(全文标明引文)
- 1 -
- 《互联网文档资源(http://www.doc88.com)》- 2019 是否引证:否
2 Movie Recommendation System based on knowledge Graph 0.5%(74)
- 2 -
2.2 Content-based Filtering For Movies.
................................................................................................... 5
2.3 Related Work
...............................................................................................................
........................ 6
Chapter 3 System Design and Architecture
......................................................................................................8
3.1 System Requirements and Constraints.
................................................................................................. 8
3.2 System Architecture Overview
............................................................................................................ 8
知
3.3 Data Sources and Processing
...............................................................................................................
网
9
3.4 Recommendation Engine
...............................................................................................................
学
.......9
3.5 User Interface and User
术
Management................................................................................................ 10
3.6 Technology Stack and Tools
..............................................................................................................
不
11
Chapter 4 Implementation and Evaluation
..................................................................................................... 12
端
4.1 Data Retrieval and Preprocessing
...................................................................................................... 12
文
4.2 Recommendation Engine Implementation
......................................................................................... 16
4.3 User Interface Development
献
..............................................................................................................
19
4.4 User Interface Running and Recommendations Testing
检
..................................................................... 23
4.5 Comparison and Discussion
测
..............................................................................................................
28
Chapter 5 Conclusion and Future Work
系
........................................................................................................ 29
5.1 Summary and Conclusions
统
...............................................................................................................
. 29
5.2 Limitations and Challenges
...............................................................................................................
29
5.3 Future Work and Recommendations
................................................................................................... 29
5.4 Implications and Impact
...............................................................................................................
..... 30
5.5 Concluding Remarks
...............................................................................................................
.......... 30
References
...............................................................................................................
...................................... 32
Acknowledgments.
...............................................................................................................
......................... 34 1
HANGZHOU NORMAL UNIVERSITY THESIS MOVIE RECOMMENDATION SYSTEM
Chapter 1 Introduction
1. 1 Background and Motivation
- 3 -
Nowadays in the era of digital people frequently get overbound by abundant content, particularly in
entertainment. Such as the platforms like Netflix, Amazon Prime, and Hulu are the main media sources
with immense libraries containing several movies and TV shows. It happens that among this stream of rich
content, the users constantly need help finding the content they might find like it. Recommendation
systems become key right here.
Selling systems are the items constructed to insert information about a certain product or a service that
matches the users’ tastes and behavior. They are now fundamental ingredients of the interface of many
online platforms that provide outstanding navigation through the complexity of choices, lending great
value to their overall user experience.
Nevertheless, one can state that although suggestions of traditional recommendation systems are a sign
of a new era in shopping, they have also inherent limitations. Often, their most important factor is that
知
they
use user ratings as the primary foundation for giving recommendations. While ratings solely explain the
网
extent to which a user likes a movie, a thorough analysis of all the data that the platform collects can
help
determine a user’s viability of a particular movie at a higher level. Two people might very well rate a
学
movie with the same numeric number for completely different reasons, or write their impressions about the
movie in their review without summing up their sentiment in a rating.
术
For this undertaking, we propose to design a method predicated on a content-based movie
recommendation system. The machine learning algorithms used by this movie reommendation system will
allow showcasing movies to users through their User Profile which was created based on their preferences
不
and past interactions. This should yield more precise and personalized movie recommendations. These
developments have the potential to bring about a new level of involvement of the users in the film
platforms; it simplifies the process of finding new movies and makes it more fun as compared to what it
端
used to be before.
1. 2 Project Overview
文
This task is all about building a custom movie advice system that takes a user's preferences into
account and augments the user's experience at movie sites. The system, whose filtering method is based on
content, compares its content with existing profiles and then generates recommendations according to the
献
similarities.
Movie content may be the genre the movie is in, the director, the cast, and also the plot of the movie.
The user profile is built up by the system, using the user's data about movies that he/she has viewed,
检
rated,
or reviewed in the past. It calculates the score of a similar movie from the content of the movie to the
测
user
profile. At this point, the movies are lined up together by the similarity score and the most similar movie
will be recommended to the user. 2
系
HANGZHOU NORMAL UNIVERSITY THESIS MOVIE RECOMMENDATION SYSTEM
The project will have a few prominent features, such as data collection and refurbishment, feature
统
reduction or extraction, recommendation algorithm implementation, and interface usability. Each element
discharges a vital function, together with what makes a complete appearance of quite a performance and
user interface.
This project has become a pivotal component as it has been a breakthrough in the recommendation
systems field. It informs us that content-based filtering can play an important role in the achievement of
accurate and relevant movie recommendation services. It not only provides the practical remedy to that
hustle of selecting a new and exciting movie from the ocean of such titles, but it also gives you a good
sense of what your neighbors and other viewers are enjoying.
1. 3 Research Questions and Scope
The project seeks to address several research questions related to content-based movie recommendation
systems:
(1) What makes the content-powered recommendation system the right one to advise a user on what to
watch depending on his or her taste?
(2) What do the movies have in their compositions that can affect the user's preferences, and how can
those be made use of in a suitable recommendation system?
(3) How exactly will the system handle new users, or newly added movies, for which there is no
historical data?
The scope of this project encompasses the development of a solution for a content-based movie
recommendation system, covering both design and operation. This involves various crucial components,
including data collection and pre-processing, feature selection, algorithm implementation, and widget
writing. The project focuses specifically on one type of recommendation system, such as
neighborhood-based or content-based systems, assuming user expectations' sincerity as essential for
- 4 -
effective and accurate recommendations. Moreover, it confines itself to system implementation, excluding
personalization or integration with the existing movie platforms(websites) and several databases, leaving
potential avenues for future research and development open. ultimately, the project aims to explore the
feasibility of content-based filtering in movie recommendations, aiming to create a tool facilitating the
discovery of new and captivating movies from a vast array of options while providing valuable academic
research data for recommendation systems.
1. 4 Significance and Contributions
The project in this area of film recommendation systems helps to establish the idea that the matter of
the film and the unique user's likes and dislikes are fundamentally important in generating movie
recommendations that the users might like. The individualized nature of such platforms goes a long way in
increasing the enjoyability of movie-hunting activities by ensuring that users do not get tired of
知
exploring
new movies due to the ease with which they find what they might like on such platforms.
网
Also, the project is making a difference not only in the field of recommendation engines but also in
demonstrating the merit of content-based filtering techniques. This algorithm review reveals the power of
elements like genres, directors, actors, a n d plot summaries in the computation for the similarity 3
学
HANGZHOU NORMAL UNIVERSITY THESIS MOVIE RECOMMENDATION SYSTEM
between films, so the recommendations are personalized and wide-range. These aren't forgetting the
术
hurdles that are faced by recommendation systems including the cold start problem subjecting new users to
their initial preference and suggesting new movies solely based on their content similarity to existing
movies.
不
Achievements at the end of the project are a huge contribution to the field, as the system of
recommendations eventually emerges as a practical solution for the problem of finding fresh and
suggestive movies among the abundance of options which also contributes to the research on
端
recommendation systems.
1.5 Thesis Structure
文
The thesis is organized into the following chapters: The thesis is organized into the following chapters:
Chapter 1: An introduction, the background and motivation of the project will be explained, followed by
an overview of the problem being studied, the research questions, and the scope of the research. The
献
significance and contribution of the work, in addition, to the thesis structure would also be stated to
provide
a clear understanding of the work being undertaken.
检
Chapter 2: This part, contains a comprehensive outlining of the literature that exists on
recommendation systems and content-based filtering. It covers the different types of recommendation
测
systems, where different recommendation systems have been used, and how they work. This also covers
the mechanism of content-based filtering and the implication of this in designing movie recommendation
systems. The chapter then goes on to critically examine what already exists by way of field research and
系
studies on movie recommendation systems.
Chapter 3: We will describe the system design and implementation aspects of the content-based film
统
attending guide system. It describes the system's needs and limitations, also, there is an overall system
design, data sources, and processing methods explanation, as well as the need and design of the
recommendation engine and user interface are mentioned here. It also discusses the system architecture and
tools applied when working on the project. -
Chapter 4: we will get into demonstrate the implementation process and its evaluation. It provides a
detailed explanation of three major technologies, namely data retrieval and preprocessing, and
recommendation engine implementation. In addition, user interface development and system evaluation
and testing are also included. Besides, the comparison between the output of the suggested system and
baseline methods or conventional recommendation systems is also provided and the strengths, weaknesses,
and prospects for attention by the author are also highlighted.
Chapter 5: The last part of the thesis which deals with a summary of the project with its key outcomes
and conclusions. To recognize any disparities and hurdles in the development and re-evaluation of the
system, it proposes directions for future investigation and advances of the system. It further explains
what
this new system will mean for the film industry, the experience of moviegoers, and other domains. 4
HANGZHOU NORMAL UNIVERSITY THESIS MOVIE RECOMMENDATION SYSTEM
Chapter 2 Literature Review
2. 1 Recommendation Systems
Recommendation engines seem to be a key element of various virtual environments that help users
conquer the vast choice options and thus improve their experience. They do a good job of determining the
opportunities that the users will have for a product or service by their probability ratings. They have a
great
- 5 -
impact on different fields of business, such as online shopping, entertainment, and social networking just
to name a few.
There are primarily three types of recommendation systems: collaborative filtration, content-based
filtration, and hybrid techniques:
A collaborative filtering algorithm assumes that user preferences have some kind of trend and that the
users who agreed in the past will also agree with them in the future. Through this, the platforms employed
a technique where the behavior of a large number of users is studied with particular respect to transaction
history, ratings, selection, and purchase information. Collaborative filtering can be further divided into
two
subtypes: people-based, which is the form of measuring the same characteristics between users, and items-
based, which is the method of finding the similarities between items. One of the pluses of collaborative
知
filtering is that it can provide highly precise suggestions by involving users' common patterns of
behavior.
网
Meanwhile,
学
2. Movie Recommendation System_第2部分 总字数:18105
相似文献列表
术
去除本人文献复制比:0%(0) 去除引用文献复制比:0%(0) 文字复制比:0%(0)
不
原文内容
端
it faces a cold start problem, which is the impediment that prevents it from recommending
fresh users or goods that lack historical data.
文
Figure 1- Classification collaborative filtering algorithms
In contrast, Content-based filtering is one kind of recommendation system that acts by reviewing the
content of the items and compares it to a user profile. Each component is represented as a certain group of
献
features which can be any word or image. Such a feature of content-based filtering does not require other
users' data, but the data only from the user itself. It, in a way, does seamless recommendation but can be
problematic, whereas it has to deal with a large amount of data about every user. 5
检
HANGZHOU NORMAL UNIVERSITY THESIS MOVIE RECOMMENDATION SYSTEM
Figure 2: Content-Based Filtering for Recommendation
测
Hybrid methods are the amalgamation of collaborative filtering and content-based filtering that permits
the two to fully manifest their unique capabilities. They can do so by rating individuals' choices and
behaviors with machine learning methods and examining the nature of the stuff as well. Yet a hybrid
系
strategy is sometimes far from being simple to put into practice and it is difficult to fit the exact
balance
between the two techniques.
统
Figure 3: Hybrid Recommender System
2. 2 Content-based Filtering For Movie
Content-based filtering is an approach to recommendation systems that recommend things by
comparing the content of the things to the user’s profile. The information of every entity is to be
represented by a set of descriptors, like words, attribute values, or images.
In the context of movies, the content can include various types of metadata, such as:
Plot summaries: The plot summary of a movie gives a summary of the narrative of the movie. It
can help provide information for content-based filtering because it contains many words
describing different movies.
Genres: The movie genre (subject). g. , action, comedy (drama) is one of the main characteristics 6
HANGZHOU NORMAL UNIVERSITY THESIS MOVIE RECOMMENDATION SYSTEM
used in content-based filtering. People often have preferences for particular genres, including
those movies so similar to them makes their suggestions closer to the truth.
Cast and crew information: The starring cast (actor, actress) and crew (director, writer, etc. ) can
also be equally important matters. The viewers could choose films with their favorite actors or
which were directed by their favorite directors.
Keywords and tags: Keywords and tags function as a source of supplemental information
regarding the elements of a film. The list may consist of motifs, locations, storylines, and other
distinguishing features of a movie.
Content-based filtering of movies has different advantages. It is free from the interference of other
users’ data since only a user's data is needed. It can suggest novel and exclusive things that match the
- 6 -
user's
unique style and preferences. But at the same time, it is not perfect. It is likely to under-emphasize the
dissimilar products and might take much data away from the users to construct a reliable profile of each of
them.
One of the common techniques applied in content-based filtering is cosine similarity. Cosine similarity
is a measure of similarity between two nonzero vectors belonging to an inner product space. Through
content-based filtering, it takes movie feature vectors to measure their similarity. It can be the feature
vector constructed of the different kinds of metadata mentioned previously. The cosine similarity between
two vectors is a value between negative and 1 that indicates how much the two vectors are in the same
direction. It is usually used in text analysis for determining the coherence of documents.
Figure 4: Content-based Form Movies Diagram
知
2.3 Related Work
Numerous studies have been conducted on content-based movie recommendation systems. These
网
studies have explored various aspects of these systems, such as the effectiveness of different
recommendation algorithms, the impact of different types of metadata on recommendation accuracy, and
the integration of user reviews and ratings into the recommendation process.
学
For instance, a study by Meel, Bano, Goswami, and Gupta titled "A comprehensive analysis of movie
recommendation system employing collaborative filtering” was published in Multimedia Tools and
术
Applications in 2020. This study discusses the prowess of the Collaborative Filtering (CF) algorithm and
its 7
HANGZHOU NORMAL UNIVERSITY THESIS MOVIE RECOMMENDATION SYSTEM
不
applications for Movie Recommendation Systems.
Another study by Iliopoulou, Kanavos, Ilias, Makris, and Vonitsanos titled “Improving Movie
Recommendation Systems Filtering by Exploiting User-Based Reviews and Movie Synopses” was
端
published in Mathematics in 20202. This study discusses how deep learning can be used in a multimodal
movie recommendation system. The study reported that the scoring prediction results show improved
文
accuracy after incorporating the potential features and connections in multimodal data with deep-learning
technology.
A third study by Mohmmad, Kanakam, Dadi, Shabana, and Pasha titled "Content and history-based
献
movie recommendation system" was published in AIP Conference Proceedings in 2022. This study
discusses the design and implementation of a complete movie recommendation system prototype based on
the Genre, Pearson Correlation Coefficient, Cosine Similarity, KNN-Based, Content-Based Filtering using
检
TFIDF and SVD, Collaborative Filtering using TFIDF and SVD3.
A study by Yuan, Qin, Yu, and Zhang titled “A Content-based Movie Recommendation System”
测
discusses the development of a content-based movie recommendation system that leverages movie
attributes such as genre, director, and cast to generate recommendations4.
Another study by Delimayanti, Laya, and Warsuta titled “Web-Based Movie Recommendation System
系
using Content-Based Filtering and Collaborative Filtering" was published in IEEE Xplore in 2022. This
study compares the performance of collaborative and content-based algorithms using different metadata
统
elements. 8
HANGZHOU NORMAL UNIVERSITY THESIS MOVIE RECOMMENDATION SYSTEM
Chapter 3 System Design and Architecture
3. 1 System Requirements and Limitations
This section outlines the functional and non-functional requirements of our system, as well as the
constraints and assumptions that have been made during its development. The system requirements define
what the system must present, while the constraints and assumptions provide context and boundaries for
the system's development. Understanding these requirements, constraints, and assumptions is crucial for
ensuring that our system is meeting the needs of the users.
Functional Requirements:
The system should, of course, have the capability to retrieve movie data from the TMDB and
CSV files.
There is a need for the informatics system to store user ratings and preferences in a database.
The system must design an intuitive interface on which the users can enter their preferences and
then view the recommendations.
Non-Functional Requirements:
The system must answer the user's requests between 2 and 10 seconds.
The system should be able to scale to at least 1000 users and 5000 film titles.
The system should be secure enough and guarantee user data privacy.
Constraints:
The system should be constructed in Python, Flask, and TMDB API.
The system will have to store data in the RDBMS (relational database management system).
- 7 -
The system has to be developed within the 12-week time frame.
Assumptions:
The TMDB API will be available to users and will be responsive.
Users will have a legitimate account and indicate ratings and preferences that can be trusted.
The system will possess sufficient computing power and the necessary storage space.
By documenting the system requirements and constraints we take into account all important functional
and non-functional aspects, design the system accordingly, and ensure its scalability, maintainability, and
security. The current section aims to present the groundwork for the following sections that will address
the design, architecture, and implementation issues.
3. 2 System Architecture Overview
Our system is designed as a microservices-based architecture, with each component serving a specific
知
purpose. The system consists of the following components:
(1) User Interface: A web-based interface built using HTML, CSS, and JavaScript, responsible for
网
interacting with users and collecting their preferences.
(2) Recommendation Engine: A Python-based service that analyzes user preferences and generates
personalized movie recommendations using cosine similarity.
学
(3) Data Retrieval Service: A service that retrieves movie data from the TMDB API and stores it in the
database. Additionally, it uses CSV metadata files to store and retrieve movie information. 9
术
HANGZHOU NORMAL UNIVERSITY THESIS MOVIE RECOMMENDATION SYSTEM
(4) Database: A relational database management system (RDBMS) that stores user preferences, movie
data, and recommendation results. Specifically, the system uses SQLite as the database management
不
system.
3.3 Data Sources and Processing
The movie recommendation system uses a combination of data sources and processing techniques to
端
provide personalized movie recommendations.
Data Sources:
文
Tmbd_5000_movies.csv, main_data.csv: Additional movie metadata, such as director and cast
information, is stored in the tmbd_5000_movies.csv, main_data.csv file.
Tmbd_5000_credits.csv: movies credit data is stored in the tmbd_5000_credits.csv file.
献
Data Processing:
The data processing component of the system is responsible for retrieving, cleaning, and transforming
the data into a suitable format for the recommendation algorithms. This involves several steps:
检
(1) Data Retrieval: This step involves fetching the movie metadata from the data sources. This could be
done through API calls or by downloading and importing data files.
测
(2) Data Cleaning: Once the data is retrieved, it needs to be cleaned. This could involve removing
duplicates, handling missing values, and correcting inconsistent or incorrect data.
(3) Feature Extraction: This step involves identifying and extracting relevant features from the movie
系
metadata. These features, such as genres, director, and cast, will be used by the recommendation
algorithm to calculate the similarity between movies.
统
(4) Data Transformation: The extracted features need to be transformed into a format that can be processed
by the recommendation algorithm. This could involve encoding categorical features, normalizing
numerical features, and vectorizing textual.
Figure 5: Data Flow Diagram
4 . 3 Recommendation Engine
The Recommendation Engine is the core component of the system, responsible for generating
personalized movie recommendations for users. It uses a content-based filtering approach to recommend
movies that are similar to the ones a user has liked or interacted with in the past. 10
HANGZHOU NORMAL UNIVERSITY THESIS MOVIE RECOMMENDATION SYSTEM
Algorithm:
Data Preprocessing: The engine preprocesses the movie metadata and user interaction data to
create a matrix of user-movie interactions.
Vectorization: The engine converts the preprocessed data into vector representations using
techniques such as bag-of-words.
Similarity Calculation: The engine calculates the similarity between movies based on their vector
representations using cosine similarity.
Recommendation Generation: The engine generates a list of personalized movie
recommendations for each user based on their past interactions and the similarities between
movies.
Components:
Movie Embeddings: This component is responsible for generating vector representations of
movies based on their metadata.
- 8 -
User Embeddings: This component is responsible for generating vector representations of users
based on their interaction history.
Similarity Matrix: This component is responsible for calculating the similarity between movies
and users.
Recommendation Generator: This component is responsible for generating personalized movie
recommendations for each user.
Figure 6: Recommendation Engine Workflow
3. 5 User Interface and User Management
The user interface and user management are vital components of the content-based movie
recommendation system. They interact with the users, and they manage the interactions that the users have
with the system.
知
User Interface: 11
HANGZHOU NORMAL UNIVERSITY THESIS MOVIE RECOMMENDATION SYSTEM
网
A user interface is the front-end element that creates the interaction with the end users. It is
a user-friendly and easy-to-operate interface for the interactions between the system and the users.
Log in and registration page: before showing the homepage, the user will have to log in.
学
Homepage: accommodates a search section and a dropdown that lists all the films.
Recommended movie details page: includes all the data on the suggested movie in the list
术
in addition to the cast crew.
More related recommendations page: right above the whole movie details page.
Our system will also have navigation features which include browsing a wide variety of movies
不
and managing user profiles.
User Management:
The user management feature executes user management functions such as user registration, login, and
端
profile management. It keeps a user profile per registered user and changes the record of the user's
activities on the system. The user profile contains data such as the user's history, likes, and system
文
interactions.
3.6 Technology Stack and Tools
The setting up of this sort of content-based movie recommendation system will be done through the
献
making use of diverse technologies and tools. The list covers such areas as programming languages,
frameworks, libraries, data processing tools, machine learning procedures, and web development items.
Programming Languages: Python is the primary language that we use in this project. Python is a
检
language with vast capability that spreads not only in data sciences and machine learning but also
in other fields. It is as simple as a procedure, has a larger library that contains a standard library,
测
and has a variety of third-party libraries that can be used for various tasks.
Frameworks: Flask, a lightweight web framework for Python, was used to create the backend of
the system. Flask has an intuitive syntax and provides developers with a high degree of
系
convenience by being lightweight and effortlessly integrable into various project architectures -
which makes it a suitable choice for small to medium-sized projects.
统
Libraries: There are Python libraries for data processing and machine learning they include. The
above-mentioned technologies are pandas for data manipulation and analysis, sci-kit-learn for
machine learning,
Tools: Various technologies like Visual Studio, Jupyter Notebook, APIs for data collection as well
and data analysis tools are also incorporated into the project. 12
HANGZHOU NORMAL UNIVERSITY THESIS MOVIE RECOMMENDATION SYSTEM
Chapter 4 Implementation and Evaluation
4.1 Data Retrieval and Preprocessing
Now, let us dive into the data retrieval and processing using the two dataset files:
Figure 7: Datasets Used
Here, we will be using Numpy and pandas so let us import them:
Figure 8: Importing Libraries
Next, we are up to reading the data using the pandas library which provides the read_csv() method to
read the CSV file and convert it into the data frame :
Figure 9: Reading the Data
Let’s check what we have in our movies and credits data frames
Figure 10: View the data
From this data frame, we need genres (to recommend movies based on category), ID (to fetch images in
the web app), keywords, overview (movie description), and title.
Next, 13
HANGZHOU NORMAL UNIVERSITY THESIS MOVIE RECOMMENDATION SYSTEM
Figure 11: Getting the credits
- 9 -
Here, we’ll need the cast (actors) and crew (director).
But these are two different datasets. We can see there are 2 columns similar in both of the datasets,
‘movie_id’ and 'title'. We need to merge them to ease the model-making.
Let’s merge them based on ‘title’ and see if it can be merged :
Figure 12: Merging the datasets
Now there is a total of 23 columns formed by merging 19 and 3 columns on 1 column.
Here, in the data frame, we have 23 unnecessary columns. As we have discussed earlier what
columns we’ll need, we will remove the unimportant ones.
Remove unnecessary columns:
Figure 13: Remove unnecessary columns 14
HANGZHOU NORMAL UNIVERSITY THESIS MOVIE RECOMMENDATION SYSTEM
知
This is our main data. We will build our model based on these columns.
But still, for a content-based recommender, everything is needed i.e. on what basis we are building in a
网
single column. Let’s say that the column is called ‘tags’. Hence, we need to convert the ‘overview’,
'genres',
'keywords', 'cast', and 'crew' in one single column as 'tags'.
学
However, the datatypes of these columns are not similar to each other. Then, first, we need to convert
them into a text and then merge these texts all together.
术
Let’s do it,
Figure 14: Applied cod to convert
By applying this simple Python code, we are iterating over each dictionary and extracting only the name
不
in it.
Now apply this to whole ‘genres’.
Figure 15: Convert genres to the desired format
端
We’ll apply the same function for ‘keywords’.
Figure 16: Convert genres to the desired format
文
Now, we are going to preprocess ‘cast’
Figure 17: Function used to convert
And here it is after preprocessing it: 15
献
HANGZHOU NORMAL UNIVERSITY THESIS MOVIE RECOMMENDATION SYSTEM
Figure 18: After preprocessing cast
After that, we are moving forward to preprocess ‘crew’:
检
Figure 19: Applying the right function to preprocess crew
The last thing we will preprocess is ‘overview’:
测
'Overview' is a string and all other columns are lists. Hence, we will convert the string to the list. How
do
we do that?
系
统
3. Movie Recommendation System_第3部分 总字数:16842
相似文献列表
原文内容
which can easily be analyzed using sentimental analysis, to detect trends and/or
preferences.
- 14 -
Addressing Cold-Start and Sparsity Issues: The formulation of methods to deal with cold-start
problems—statistically no user preference or surveys—and data sparsity issues user reviews or metadata is
limited—could be needed to grow the accuracy of the system and make it more acceptable. Techniques
like content-based and cold-start user filtering, and, also, hybrid strategies that combine content-based
and
user past behavior algorithms could be tried.
Improving User Interface and Personalization: By habitually working on the UI and discovering
personalization methods such as adaptive interfaces or tailored content pre-severance could lift both user
experience and user engagement. Taking consumer's thoughts and tastes into consideration by incorporating
such information into the interface may in turn bring into existence an experience that is more in tune and
personalized.
知
Integrating with Streaming Platforms: Establishing integration capability with the already existing
media streaming platforms or movie datasets could result in higher utilization and effect of your system.
网
Indefinitely, platform tie-in can lead to the incorporation of additional datasets, better user bases, and
the
challenges of real-world deployments.
学
5. 4 Implications and Impact
It has the potential to impact both the field of recommendation systems as well as the movie industry
术
as a whole. The personalization of recommendations helps to improve the experience of users,
increases the engagement and retention rate of users, and ultimately leads to an increase in viewership
and revenue for movie platforms. Also, the use of similarity analysis in movie recommendation systems
不
can result in the improvement of recommendation system research and the creation of new developments
in this area. Effectively implemented AI technology in art could lead to the use of this system in other
areas like music, books, or e-commerce, thereby widening the implementation reach of this research.
端
5. 5 Concluding Remarks
In conclusion, the system represents a promising approach to movie recommendation, demonstrating
文
the potential benefits of integrating sentiment analysis into content-based recommendation systems. While
challenges remain, the system provides a solid foundation for future research and development in this area.
The lessons learned from this project, along with the proposed future directions, provide valuable insights
献
for researchers, practitioners, and stakeholders in the field of recommendation systems. As the field 31
HANGZHOU NORMAL UNIVERSITY THESIS MOVIE RECOMMENDATION SYSTEM
continues to evolve, the system serves as a testament to the potential of combining traditional
检
recommendation techniques with sentiment analysis to enhance the quality of personalized
recommendations. 32
测
HANGZHOU NORMAL UNIVERSITY THESIS MOVIE RECOMMENDATION SYSTEM
References:
(1) Adomavicius, G., & Tuzhilin, A. (2005). Toward the next generation of recommender systems.
系
IEEE Transactions on Knowledge and Data Engineering, 17(6), 734-749.
(2) Pazzani, M. J., & Billsus, D. (2007). Content-based recommendation systems. In The adaptive web
统
(pp. 325-341). Springer, Berlin, Heidelberg.
(3) Lops, P., De Gemmis, M., & Semeraro, G. (2011). Content-based recommender systems: State of the
art and trends. In Recommender Systems Handbook (pp. 73-105). Springer, Boston, MA.
(4) Liu, B. (2012). Sentiment analysis and opinion mining. Synthesis lectures on human language
technologies, 5(1), 1-167.
https://www.morganclaypool.com/doi/abs/10.2200/S00416ED1V01Y201204HLT016
(5) Pang, B., & Lee, L. (2008). Opinion mining and sentiment analysis. .
https://www.nowpublishers.com/article/Details/INR-011
(6) Musto, C., Semeraro, G., Lops, P., & Gemmis, M. (2017). Generating computer programming
exercises from user interactions with MOOCs in a context-aware perspective. Applied Intelligence,
47(1), 3-19.
(7) Harper, F. M., & Konstan, J. A. (2015). The MovieLens datasets: History and context. ACM
Transactions on Interactive Intelligent Systems (TiiS), 5(4), 1-19.
(8) Pazzani, M. J. (1999). A framework for collaborative, content-based, and demographic filtering.
Artificial Intelligence Review, 13(5-6), 393-408.
(9) Pazzani, M., & Billsus, D. (1997). The identification of interesting websites. Machine Learning,
27(3), 313-331.
(10) Melville, P., Mooney, R. J., & Nagarajan, R. (2002). Content-boosted collaborative filtering for
improved recommendations. Aaai/iaai, 23, 187-192.
(11) Tintarev, N., & Masthoff, J. (2007). A survey of explanations in recommender systems. (pp.
801-810). IEEE.
(12) Beel, J., Gipp, B., Langer, S., & Breitinger, C. (2016). Research-paper recommender systems: a
- 15 -
literature survey. International Journal on Digital Libraries, 17(4), 305-338.
(13) Bobadilla, J., Ortega, F., Hernando, A., & Gutiérrez, A. (2013). Recommender systems survey.
Knowledge-based systems, 46, 109-132.
(14) Ricci, F., Rokach, L., & Shapira, B. (2011). Introduction to recommender systems handbook. In
Recommender Systems Handbook (pp. 1-35). Springer, Boston, MA.
(15) Sayedelahl, A., Dumitrache, A., Balters, S., & Koenen, J. (2019). Movies Recommendation Using
Sentiment Analysis from Microblog, Metadata, and Interdependency between Ratings. arXiv
preprint arXiv:1908.07172. 33
HANGZHOU NORMAL UNIVERSITY THESIS MOVIE RECOMMENDATION SYSTEM
(16) Singh, V. K., Piryani, R., Uddin, A., & Waila, P. (2013). A new feature-based heuristic for
aspect-level sentiment classification. (pp. 711-717). IEEE.
知
(17) Killian Duay, Personalized web-based application for movie recommendations
(18) Wang, K., Poo, D. C., Walker, E. A., & Yu, J. (2020). An Integrated Machine Learning Model for
网
Sentiment Analysis and Topic Classification. 34
HANGZHOU NORMAL UNIVERSITY THESIS MOVIE RECOMMENDATION SYSTEM
Acknowledgments
学
I would like to thank Hangzhou Normal University - School of Information Science and Technology for its
help and provision of all the resources that are needed for the completion of my project called "Movie
术
Recommender System".
I shall never forget what my supervisor (Fanran 范然) has done not only with the kindness of sharing with
me her experience and knowledge but with great encouragement and support too. By advancing this effort
不
as well as her acumen, he played a vital role in fully embracing this project.
Besides, I would like to thank a lot all the faculty members and staff of the School of Information Science
and Technology and also express my gratitude for their assistance and help during the implementation of
端
this project
Furthermore, I must mention the extensive help and enthusiasm of my family members and friends whose
文
loyalty to me is the well-spring of my hope and thinking.
献
说明:1.总文字复制比:被检测论文总重合字数在总字数中所占的比例
2.去除引用文献复制比:去除系统识别为引用的文献后,计算出来的重合字数在总字数中所占的比例
检
3.去除本人文献复制比:去除作者本人文献后,计算出来的重合字数在总字数中所占的比例
4.单篇最大文字复制比:被检测文献与所有相似文献比对后,重合字数占总字数的比例最大的那一篇文献的文字复制比
测
5.复制比:按照“四舍五入”规则,保留1位小数
系
6.指标是由系统根据《学术论文不端行为的界定标准》自动生成的
7.红色文字表示文字复制部分;绿色文字表示引用部分(包括系统自动识别为引用的部分);棕灰色文字表示系统依据作者
统
姓名识别的本人其他文献部分
8.本报告单仅对您所选择的比对时间范围、资源范围内的检测结果负责
- 16 -