NATIONAL UNIVERSITY HOCHIMINH CITY UNIVERSITY OF INFORMATION TECHNOLOGY ADVANCED PROGRAM IN INFORMATION SYSTEMS Nguyen Hoang Long — 19521787 Nguyen Trung Hieu — 19521507 BACHELOR OF ENGINEERING IN INFORMATION SYSTEMS THESIS ADVISORS Dr. Do Trong Hop Dr. Tran Van Thanh HO CHI MINH CITY, 2024 ii ASSESSMENT COMMITTEE The Assessment Committee is established under the Decision. by Rector of the University of Information Technology.
- Member ACKNOWLEDGMENTS After a period of research and writing my graduation thesis, we have basically completed the set goals. Getting those results is due to my own efforts and efforts, along with the encouragement of my family, friends and teachers. That is a great encouragement to help me complete this graduation thesis well. We would like to express my deep gratitude to Dr.
Do Trong Hop and Dr Tran Van Thanh directly guided and enthusiastically helped me during the completion of this graduation thesis. We would like to express my gratitude to the teachers of the University of Information Technology, especially the teachers of the Information Systems Department, for giving me the opportunity to study and absorb knowledge so that we can Complete this thesis and prepare a good foundation for future work. We would also like to thank all our friends who have directly and indirectly helped us in doing this thesis. Last but not the least, we place a deep sense of appreciation to our family members who have been constant source of inspiration for us.
Any kind of suggestion or criticism will be highly appreciated and acknowledged. Table of Contents ACKNOWLEDGMENTTS.s- <5 (G00 005004088066 1 Table of Content .010404 0150090098090 890 2 LIST OF EIGUIRES.1 Šfa†€IN€HI. Ăn TH HH HH HH 7 1. What is Music Recommendation System? .3 Objectives Of ThheSIS.
5 nh ng gưkt 8 1.- --- cv n9 HT Hà HH Hành 9 1.5 Scope of the the€SI1S. Overview Algorithm Research .1 Historical of Music Recommendation .2 Application of music recommendation system.3 Common filtering methods used in Music Recommendation Systems.1 Content-Based Filtering: .--- ¿+ ss++ex++sexeexeereereeeeses 16 2.4 Specific Challenges of Music Recommendation.5 Current situation of Music Recommendation system. Methods Of Algorithm Implemenfation.2 Content-based EIÏt€rIng.1 Content-based Filtering Introduction.2 Content-based Filtering in Music Recommendation Systems.3 Collaborative FIÏt€TINE.- -- -- 5 S5 k1 k9 vn ng ng rưn 29 3.1 Collaborative Filtering Introduction.2 User-ltem điafAS€K.- 5 cv 1H ng ri 30 3. Memory-based CF Techn1ques.4 Memory-based CF Techn1ques.5 Item-based Neighborhood.
--ss«ss+sx+sxsessseseesxss 34 3.6 Model-based CF Techniques .7 Hybrid Memory- and Model-based Techniques .8 Collaborative Filtering in Music Recommendation Systems.4 Conclusion from the reSuÏ(. Analysis and Design The Applicafion.---- 5 6 xà kh TH HT HH 48 4.5 Sequence Diagram nan.1 1211 H1 1111111 1 1n nh Thun ngàng 52 5. ch HH HH rệt 52 5.- Gà ng gi, 53 695. ame Le ys ssssssesscrdecsssesseghecrsorsorsenseseeees 53 5.
HH cọ TH HT 0009 0890000 55 APPENDICEG 100. "nh 56 56 LIST OF FIGURES & Figure 3.3 TF-IDF FOrmUÌa. Cosine Similarity measure 1llusfrafIOn. Result of Content-based Filtering.
User-item matrix for collaborative filtering. Categories rating điaðTATH.-- - 5s x se kesersererree 40 Figure 3. Evaluating RMSE of algorithm SVD. Result of Collaborative FIltering.
icc eseeeesecreeseeecesseeeeseeseeseeeaes 45 Figure 3. Use Case Diagram of the Sysfem. R diagram of the System. cc eccscseeeseeeeeseeseeseeseeeeeneeeeeees 48 Figure 3.
Sequence Diagram of the System. ccc ceccesceseesessecseeeeceseeseeseceecsteeseeseeaeeaeenseeeeees 56 Appendices ScreenShot 2.- -- -- c2 212111 1151111191111 1111111 11g rệt 56 LIST OF TABLES & Table 1. Examples of Music Recommendation ABSTRACT Music is an indispensable part of our lives. We listen to music every day according to our preferences and mood.
With the advancement and increase in the amount of digital content, people's choice to listen to diverse music genres has also increased significantly. Therefore, the necessity of delivering the most suitable music to the listener is an interesting area of research in computer science. The music recommendation system aims to reduce the task of finding music items that might interest users by creating meaningful content Recommendations. Music recommendations are different from book recommendations and film, due to its low cost per item, short consumption time, high reusability per item, contextual usage and variety of items.
Understand music listening styles and consumption is important to create accurate and satisfying music recommendations. One of the important measures to bring the best music to listeners is each person's personality traits. In this project, we aim to explore the impact of personality traits Regarding collaborative filtering (user to user), this is one of the most popular proposals engines used today. To determine a person's personality, we will rely on data from the number of times a song is listened to, a person's favorite music genre, and then make recommendations through user collaborative filtering techniques to Music recommendation.1 Statement Music is an essential part of human life.
Music is the pleasant sound that leads us to experience harmony and higher happiness. With the advancement in technology, music has significantly progressed and increased in terms of quality and volume. The type of music people created and listen differs according to place and culture. The taste of music even differs from person to person and even in moods of same person.
So, it is very useful if we could determine some method to find what kind of music a person might be interested in listening and use this finding to recommend music to him/her. And music recommendation system can help users discover new songs: With today's huge music store, finding new songs that suit users' interests is a challenge. Music recommendation system can help users discover new songs based on songs they have heard before.2 What is Music Recommendation System? Music recommendation system is system help consumers and the music industry with the discovery and delivery of music. In order to realize the personalized distribution of music, it may be beneficial for recommender designers to understand the music listening behaviors and know about the state of music consumption in the industry.
Understanding user preference and behavior can help to propose a reasonable recommendation to a specific user. For example, some users show a clear bias towards style when choosing music, while some emphasize timbral similar. In order to make recommendations respectively to these two types of listeners, the recommender needs to focus on different attributes (e., musical style and timbre). Moreover, user feelings and expressions can be different towards the same music, such that a personalized user profile is needed for each user before the system can make meaningful recommendations.
Generally, a user’s preference shifts with time, in terms of years, seasons, days, and even hours. For instance, a user who liked calm and soft music before, may like noisy music now. So a user’s profile needs update and maintenance to describe the music preference of the user at a time. Unlike the consumption of movie, books, and games, people listen to music repeatedly and continuously.
This adds more complexity to capture a user’s preference accurately, which is important for a music recommendation system. In reality, the frequency distribution of music transactions is concentrated at the beginning (high volume); dominated by a very few popular items (the well-known hits) followed by a long list of items that does not sell that well. This long list is referred as the Long Tail, where many songs or artists, the misses, are only played or downloaded by a relatively small group of people. It seems that most users prefer popular titles, while only a minority exploit niche titles.
Analyses of the music sales have shown that the music industry is dominated by popular artists and songs. However, nearly everybody’s taste deviates from the mainstream somewhere, with most people consuming niche products at least some of the time. Sometimes, listeners may be expecting to discover and enjoy a wide range of music that may be less popular but a good match to their personal taste. On the other hand, some economists explored the utility of the tail items and proposed the Long Tail business strategy.
That is, offering customization to individual consumers can increase the profit in e-commerce by “selling less of more”. Both the users and the profit-driven service providers are looking forward to advanced tools for music discovery and recommendation.3 Objectives of Thesis 1. User Satisfaction and Engagement: - Personalization: Recommend music that users are likely to enjoy based on their individual preferences, listening history, and behavior. This helps users discover new music they might love and keeps them engaged with the platform.
- Reduce decision fatigue: Help users navigate the vast amount of music available by making it easier to find what they want to listen to. This can save users time and frustration, ultimately leading to a more positive experience. - Increase user retention: By providing users with music they enjoy and keeping them engaged, recommendation systems can help platforms retain users and prevent churn 2. Content Discovery and Exploration: - Serendipity and surprise: Recommend music that users might not have found on their own, but still find enjoyable.
This can help them expand their musical horizons and discover new artists and genres. - Diversity and balance: Recommend a variety of music, not just the most popular or similar to what the user has already listened to. This helps ensure that users are exposed to a wider range of music and prevents them from getting stuck in filter bubbles. - Trend and novelty detection: Recommend new and trending music to keep users up- to-date and engaged with the latest musical landscape.4 Research Method Their method of researching music recommendation systems is to collect data on the number of times a user has listened to a song from a list of thousands of songs that user has listened to, then line up and filter out the songs and genres that the user has listened to.
listened and then based on the results to recommend music according to style, genre, and singer that has similarities with that user's playlist. Nowadays, digital data on the Internet has been massive than ever, which have created a potential challenge of information overload, hindering timely access of items of interest on the Internet 1.5 Scope of the thesis Thesis Scope: Research thesis can focus on one or more specific goals, such as: e Build a new music recommendation system with new features and algorithms. e Improve the performance of existing music recommendation systems. e Research new issues in the field of music recommendation systems.
Content of the thesis: The research project may include the following contents: e Overview of music recommendation systems. e Research music recommendation methods and algorithms. e Collect and process data. e Build and evaluate recommendation models.
e Propose further research directions. Research subjects: Research thesis can focus on one or more specific subjects, such as: e Music recommendation system in a specific music website. e Music recommendation system for a specific group of users, such as young users, middle-aged users, elderly users,. e Music recommendation system for a specific music genre, such as pop music, rock music, classical music,.
Research methods: Research projects can use different research methods, such as: e Theoretical research.6 Outline This thesis identifies issues that need to be addressed to develop music recommendations systems and how they have been handled so far. By studying history develop music recommendation methods, theses enable designers and researchers Use accumulated knowledge and experience addresses many open questions that require further exploration. The remainder of this thesis is organized as follows: Chapter 2 describes about Music Recommendation and the Reality of today's music market. Chapter 3 describes how collaborative filtering and content-based filtering works and reviews its application in Music suggestion system.
Collaborative filtering depends on a set of human evaluations (called ratings) of items to predict how much a user will like an item that doesn't. Rating data can be collected explicitly or implicitly. Memory-based and Model-based methods have their own advantages and disadvantages. Although collaborative filtering is a successful recommendation method, it poses some challenges such as data sparsity, common trends, and cold starts.