VIETNAM NATIONAL UNIVERSITY HO CHI MINH CITY HO CHI MINH CITY UNIVERSITY OF TECHNOLOGY NGUYEN QUYNH ANH PHUONG SESSION-BASED RECOMMENDATION SYSTEM IN FASHION Major: Computer Science Major code: 8480101 MASTER’S THESIS HO CHI MINH CITY, June 2024 THIS THESIS IS COMPLETED AT HO CHI MINH CITY UNIVERSITY OF TECHNOLOGY – VNU-HCM Supervisor(s): • Assoc. Thoai Nam, Ph.D • Nguyen Quang Hung, Ph. Huynh Tuong Nguyen, Ph.D Examiner 2: Ha Viet Uyen Synh, Ph.D This master’s thesis is defended at HCM City University of Technology, VNU- HCM City on 18th June 2024 Master’s Thesis Committee: 1. Chairman: Le Thanh Sach, Ph.
Secretary: Le Thanh Van, Ph. Huynh Tuong Nguyen, Ph. Examiner 2: Ha Viet Uyen Synh, Ph. Commissioner: Nguyen Le Duy Lai, Ph.D Approval of the Chairman of Master’s Thesis Committee and Dean of Faculty of Computer Science and Engineering after the thesis is corrected (If any).
CHAIRMAN OF THESIS COMMITTEE DEAN OF FACULTY OF COMPUTER SCIENCE AND ENGINEERING i VIETNAM NATIONAL UNIVERSITY - HO CHI MINH CITY SOCIALIST REPUBLIC OF VIETNAM HO CHI MINH CITY UNIVERSITY OF TECHNOLOGY Independence – Freedom - Happiness THE TASK SHEET OF MASTER’S THESIS Full name: Nguyen Quynh Anh Phuong Student ID: 2171069 Date of birth: 02/12/1996 Place of birth: Lam Dong Province Major: Computer Science Major ID: 8480101 I. THESIS TITLE: Session-based recommendation system in Fashion (Hệ thống gợi ý dựa vào phiên làm việc ngành thời trang) II. TASKS AND CONTENTS: • Task 1: Research and experiment for Data Augmentation techniques in preprocessing data of session-based recommendation system to create more variations of input data. • Task 2: Research and proposed suitable approaches for predicting next items in session- based recommendation system domain Fashion using deep neural networks approaches.
• Task 3: Experiment and evaluate proposed approaches. THESIS START DAY: 15/01/2024 IV. THESIS COMPLETION DAY: 20/05/2024 V. Thoai Nam, PhD and Nguyen Quang Hung, Ph.D Ho Chi Minh City, 15/08/2024 SUPERVISOR 1 SUPERVISOR 2 CHAIR OF PROGRAM COMMITTEE (Full name and signature) (Full name and signature) (Full name and signature) DEAN OF FACULTY OF COMPUTER SCIENCE AND ENGINEERING (Full name and signature) ii ACKNOWLEDGMENT First and foremost, I would like to express my deepest appreciation to my beloved parents, my family and especially my aunt Mrs.
Nguyen Thi Hanh for inspiring and supporting me throughout my academic and professional endeavors. Furthermore, I would like to thank Ho Chi Minh City University of Technology gave me an invaluable opportunity to pursue my academic aspirations. The journey to learn and grow with a support and guidance from great teachers and colleagues is the best transformation to cultivate my character and knowledge. I want to express my profound appreciation to Associate Professor Thoại Nam of Computer Science and Engineering at the Ho Chi Minh City University of Technol- ogy, for his supportive guidance, encouragement, and invaluable feedback throughout this study.
I am immensely grateful for his patience in guiding me and reviewing my work. Without his guidance and support, this thesis would not become possible. I also would like to thank Mr. Nguyen Tan Sang, Mr.
Dao Ba Dat, Mr. Ha Minh Duc, Ms. Vo Thi Kim Nguyet for their enthusiastic cooperation and encouragement in offering valuable advice during the thesis. Last but not least, I would like to thank for my company and my leader for always give me a chance to still work while studying.
This experience has not only enriched my professional career development but also instilled me in pursuing knowledge. iii ABSTRACT OF DISSERTATION Recommendation system plays a key role in business and customers focus on im- proving customers shopping experience, creating customer satisfaction by understand their preferences. Recently, session-based recommendation system has been arisen to capture the short-term but dynamic user preferences. This helps business engage users quickly to explore products and make decision instead of become their loyalty customers first.
Recommendation system is widely applied in various domain includ- ing fashion. Fashion domain is challenge domain due to its fast changing trends and enormous items adapt to customer’s tastes. Deep learning based approaches are widely used in many domains and recom- mendation system. Furthermore, there are more and more approaches and techniques are applied in intersection domain between Computer Vision, Deep Learning, Natu- ral Language Processing.
These methods are applied on a problem of session-based recommendation system as well. This thesis focus on researching and conducting experiments to building recommendation system with the target to predict the next items in the customer’s sessions based on deep neural networks approaches. This thesis contributes to the Fashion recommendation system when experiment and apply techniques various techniques from other fields to solve problems. There are some highlight points to address here: • The first highlight is a research, experiments and evaluation when apply data augmentation techniques to create variance data input for session-based recom- mendation system.
• The second highlight is a research, experiments and evaluation approaches using Deep Neural Networks focused on Attention and Neural Networks to predict the next items in a anonymous sessions. Furthermore, applying some techniques to building recommendation system in Fashion domain. iv TÓM TẮT LUẬN VĂN Hệ thống gợi ý đóng vai trò quan trọng đối với cả doanh nghiệp và khách hàng. Đối với khách hàng, hệ thống gợi ý giúp cải thiện trải nghiệm mua sắm, tăng sự yêu thích của khách hàng với sản phẩm khi thể hiện sự thấu hiểu sở thích, cung cấp nhiều hàng hóa phù hợp với nhu cầu của họ.
Ngoài ra, hệ thống gợi ý theo phiên ngày càng được thu hút nghiên cứu và ứng dụng khi giúp doanh nghiệp có thể nhanh chóng thấu hiểu khách hàng, cung cấp những trải nghiệm chỉ với một vài tương tác. Điều này giúp khách hàng có thể dễ dàng sử dụng sản phẩm và không cần trở thành thành viên trước tiên nhưng vẫn có thể khám phá và mua hàng. Hệ thống gợi ý được áp dụng rộng rãi với nhiều ngành kể cả ngành thời trang với nhiều thách thức như thay đổi xu hướng nhanh chóng, số lượng sản phẩm đa dạng, nhu cầu cá nhân hóa. Những năm gần đây, những giải pháp dựa trên Học sâu đang được áp dụng rộng rãi cho mảng hệ thống gợi ý.
Hơn thế nữa, nhiều giải pháp kĩ thuật lai giữa các mảng Học sâu, Xử lý ảnh và Xử lý ngôn ngữ tự nhiên cũng đang được áp dụng rộng rãi với nhiều ứng dụng đặc biệt hệ thống gợi ý theo phiên. Bài luận văn này sẽ chú trọng đến nghiên cứu, thí nghiệm và đánh giá khi áp dụng các giải pháp trên để xây dựng hệ thống gợi ý với mục tiêu có thể dự đoán được những sản phẩm tiếp theo trong phiên hoạt động của khách hàng. Những nghiên cứu này đóng góp khi áp dụng cho ngành thời trang và hệ thống gợi ý khi thí nghiệm, đánh giá khi áp dụng nhiều kĩ thuật áp dụng khác nhau. Một vài điểm nhấn được liệt kê như sau: • Đầu tiên, nghiên cứu, thí nghiệm và áp dụng kĩ thuật làm giàu dữ liệu trong giai đoạn tiền xử lý dữ liệu - được sử dụng rộng rãi với những ngành khác như Xử lý ảnh, Xử lý ngôn ngữ tự nhiên cũng sẽ được áp dụng trong bài toán hệ thống gợi ý dựa vào phiên.
• Tiếp theo, nghiên cứu, thí nghiệm và đánh giá khi xây dựng hệ thống gợi ý dựa vào phiên sử dụng áp dụng Mạng nơ-ron sâu để dự đoán những sản phẩm tiếp theo trong phiên gợi ý ẩn danh. Áp dụng để xây dựng hệ thống gợi ý trong ngành thời trang. v DECLARATION I declare this thesis to be a work of mine under the supervision of Assoc. Thoại Nam was built to meet society’s demands, and my ability to achieve informa- tion.
The data and figures presented in this thesis for analysing, comments and eval- uations from various resources by my own work and have been duly acknowledged in the reference part. In addition, the contents of external assistance should be recorded, referenced, and cited. I will take full responsibility for any fraud detected in my thesis. Ho Chi Minh City University of Technology (HCMUT) - VNU-HCM is unrelated to any copyright infringement caused on my work (if any).
Ho Chi Minh City, June 2024 Nguyen Quynh Anh Phuong Contents List of Figures. ix List of Tables .2 Recommendation system in Fashion .4 Objectives and Missions .5 Scope of work .2 Recurrent Neural Networks .3 Gated Recurrent Unit .1 Session based Recommendation System .2 Natural Language Processing and Recommendation system relationship 29 3.1 Data Augmentation methods .2 Data augmentation applied strategies .1 Neural Attentive Session-based Recommendation .2 Behaviour Sequence Transformer .3 Bidirectional Encoder Representations from Transformer. 60 5 EXPERIMENTS AND EVALUATION 62 5.1 Experiment data augmentation .2 Experiment NARM model .3 Experiment Transformer-based model .4 Experiment BERT4Rec .3 Results and discussion .1 Data Augmentation techniques in preprocessing .2 Deep neural networks based approaches in processing. 81 List of Figures 1.1 The difference between Sequential recommendation system and Ses- sion recommendation system [2] .2 A session of anonymous user in session-based recommendation sys- tem problem description .1 Neural Networks in Deep Learning .3 Feedforward in Neural Networks in NLP [4] .4 Selecting the embedding vector for word V5 by muliplying the embed- ding matrix E with a one-hot vector with a 1 in index 5 [4] .5 Recurrent Neural Networks training [4] .6 GRU cell and it’s gates [5] .7 GRU explained with equation .9 Mechanism of First Stack of Encoder [7] .10 One-head Attention in Transformer [7] .11 Decoder in Transformer .12 SkipGram model Word2Vec [8] .1 The categorization of session-based recommendation system approaches 29 3.2 Timeline relationship of Natural Language Processing and Recom- mendation system [11] .3 Data Augmentation in text classification Natural Language Process- ing taxonomy [25] .1 Item interaction counts distribution .2 Item interaction distributed by date .3 Item interaction distributed by week .4 Item interaction distributed by month .5 Data Augmentation methods: Noise Injection, Redundancy Injection, Random Swap, Random Deletion, Synonym Replacement .6 Data Augmentation strategies applied.
Example: get fraction 20% of the dataset and using method Swap Random, Naug = 4, keep the original one and generate 3 more sequences .7 Overview framework of encoder-decoder-based NARM [17] .8 Global encoder in NARM model - the last hidden state is interpreted as the user’s sequential behaviour feature .9 Local encoder NARM .10 Model of NARM .11 Transformer-based model based on Behaviour Sequence Transformer .12 Label smoothing with ε 0.13 BERT4REC model architecture overview .14 Transformer layer in BERT .15 The non-linearity in the negative range of GELU [35] .16 BERT4Rec model by stacking many Transformer layers. 58 List of Tables 4.1 Compared Dressipi dataset in Fashion with other popular datasets .1 Parameters for NARM model .2 Parameters for Transformer model .3 Parameter for BERT4Rec .4 Compare when apply Noise Injection data augmentation method with different fraction and Naug .5 Compare when apply Redundancy Injection data augmentation method with different fraction and Naug .6 Compare when apply Random Swap data augmentation method with different fraction and Naug .7 Compare when apply Random Deletion data augmentation method with different fraction and Naug .8 Compare when apply Synonym Replacement data augmentation method with different fraction and Naug .9 Data augmentation methods applied comparison overall .10 Recall@20 metric result when experiment impact of learning rate of Optimizer Adam. Epochs: 5; batch size 512, hidden size 100, embed- ding size 50 .11 Impact of max length sequence in overall results with Recall@20 met- ric. Learning rate of Optimizer Adam is 0.01, epochs is 10, batch size is 512, hidden size 100, embedding size is 50) .12 Impact of max length sequence in Recall@20 metrics.
Experiment with learning rate of Optimizer Adam 0.