Department of Computer Science & Engineering
Hanyang University, Seoul, Korea
Sang-Wook Kim received his Ph.D. degree in Computer Science from Korea Advanced Institute of Science and Technology (KAIST) in 1994. From 1995 to 2003, he served as an Associate Professor of the Division of Information and Communications Engineering at Kangwon National University. In 2003, he joined Hanyang University, Seoul, Korea, where he currently is a Professor at the Department of Computer Science & Engineering and has been recognized as a (research) distinguished professor in 2019. He is the director of the Brain-Korea-21 research program since 2014, and also has been leading a National Research Lab Project from 2015. His research interests include databases, data mining, social network analysis, recommendation, and web data analysis.
From 2009 to 2010, Professor Kim visited the Computer Science Department at Carnegie Mellon University as a Visiting Professor. From 1999 to 2000, he worked with the IBM T. J. Watson Research Center as a Post-Doc. He also visited the Computer Science Department of Stanford University as a Visiting Researcher in 1991. He is an author of over 140 papers in refereed international journals and international conference proceedings. He served Program Committees of over 100 international conferences including ACM KDD, IEEE ICDE, IEEE ICDM, VLDB, WWW, and ACM CIKM. He is now an associate editor of two international journals: Information Sciences and Computer Science & Information Systems (ComSIS). He received the Presidential Award of Korea in 2017 for his academic achievement and he is currently a member of National Academy of Engineering of Korea from 2019. He is also a member of the ACM and the IEEE.
As the number of online items such as products, contents, and people significantly grows these days, it becomes a difficult task for users to find the items on their own. Good matching of users to their suitable items is very important to enhance users’ satisfaction and companies’ profit, which highlights the necessity of recommendation systems technology. The recommendation system analyzes the characteristics of users’ past behaviors, predicting the items with which individual users would be truly satisfied. In this talk, we discuss the concepts, techniques, and applications of recommendation systems. We start with the concepts of recommendation systems and introduce real-world applications in a variety of business fields. Then, we explain three categories of approaches to recommendation systems: content-based, collaborative-filtering-based, and trust-based approaches. Next, we describe machine learning techniques widely applied in developing recommendation systems. Finally, we share the state-of-the-art techniques developed in our group and show their effectiveness and efficiency with evaluation results. The agenda of this talk is organized as follows.
I. Introduction 1.1. Concepts 1.2. Applications 1.3. One-class and multi-class settings II. Approaches 2.1. Content-based approaches 2.3. Collaborative-filtering-based approaches 2.4. Trust-based approaches III. Techniques 3.1. Heuristic techniques 3.2. Matrix-factorization techniques 3.3. Graph-analysis techniques 3.4. Deep-learning techniques IV. State-of-the-art techniques 4.1. Data imputation techniques by exploiting uninteresting items 4.2. Graph-theoretic techniques with data imputation 4.4. Multi-type Bayesian pair-wise ranking techniques 4.5. Adversarial signed graph embedding techniques 4.6. GAN-based CF techniques V. Summary and conclusions
Computer Science
University of Eastern Finland
Pasi Fränti received his MSc and PhD degrees from the University of Turku, 1991 and 1994. Since 2000, he has been a professor of Computer Science in the University of Eastern Finland (UEF). He has published 93 journals and 173 peer review conference papers. His main research interest is in clustering and location-based services but is currently doing research also in web mining and health informatics. He likes running and travelling, and has collected all his related activities during 2009-2019 into Mopsi location-based service to provide location tagged material with ground truth.
Geolocation allows new ways to perform data analysis, search and recommendations. We can track user movements, retrieve information by location-based queries, predict user behavior and give recommendation for mobile users about activity events and for sight-seeing. This presentation gives a review of recent research results and future trends in this area. We cover processing, storage and event extraction from GPS trajectories; summarize how to build location-aware search engine and recommendation system; study content creation and analysis of user performance in location-based gaming. Methods for classifying users and their similarities based on location history is given.
For more info: http://cs.uef.fi/pages/franti/