Prof. Takahiro Hara: EAI Distinguished lecture on Cross-domain Approaches or User Activity prediction in Online and Real Worlds

Prof. Takahiro Hara from Osaka university outlines how user activities changed for both online and real world, as a part of EAI Distinguished lectures program.

In smart cities, there are various kinds of services to support people’s daily life for both online and offline. Since there are so many applications in smart cities, personalization is very important. So services should adaptively change according to user preferences and characteristics. Learn more in the video below:

Personalization: problem and the idea

It is important to provide personal services by analyzing big subs data, such as the service usage log to provide users with comfortable services. However, currently, there is a strong limitation: it is difficult to share user or personal model construct in different service providers or domains due to the published issue and interest or light issue. So the idea of this research is to transfer the prediction model between domains without exchanging raw data and user IDs.

Web advertisement & E-commerce domain

Many E-commerce services sell a large number of products in a short time. So this is kind of rush sales: many new products appear on the market, but also disappear from the market very quickly. In addition, most of the users have just a few or no experiences of real purchase, so they’re just looking for or getting some free coupon. In this sense, there are so many new users and new products – so there’s no interaction between them in terms of the purchase. That’s why it’s difficult to model the user’s purchase pattern only from the data obtained in the E-commerce domain.

Online & Real world: location service

The online and real-world needs to be integrated or transformed to perform some customer approach. So one possible approach is modeling personas or users as embedding objects. After that, the transforming the vectors between the domains should be performed. In this research, the authors take into account not only the position but also state the time and number of people for each position. That means that the authors assign some semantics or meaning to each location. They don’t directly model the user, but they first try to model each location and then model the user.

Takahiro Hara is a full professor in the Department of Multimedia Engineering, Osaka University. His research interests include database systems, distributed systems, mobile computing, and social computing. Prof. Hara took part in more than 20 EAI conferences and became EAI Fellow for his extensive involvement. Prof. Hara presented his Distinguished lecture on Cross-domain Approaches or User Activity prediction in Online and Real Worlds at SmartCity 360 2020 convention.

The honorable Distinguished lectures program serves as an inspiration featuring great talks by the world’s most prominent researchers. Let us introduce to you our featured world’s top speakers presenting a topic as our distinguished guest lecturers.

Watch EAI’s previous Prof. Paul A. Watters: EAI Distinguished Lecture on Cyber Security Knowledge, Skills and Abilities: Designing Digital Education for the 21st Century