The relational network of people (RNP) model has been attracting the interest of not only researchers but also industrial engineers. RNP can be constructed from friend lists in online social networking services (SNSs) and from inter-contact logs between individuals. One of the killer applications of RNP is the prediction of user demands, which is key to maximizing user satisfaction in content delivery services such as video streaming and video advertising. It is well known that an RNP representing social closeness between individuals (a so-called social network) can estimate user preferences simply, as we expect that people close to each other will have similar preferences. However, although there are many metrics that enable the social closeness between individuals to be measured, it is unclear which metric is best suited for individual services. Therefore, this paper introduces a new approach based on brain imaging. Brain imaging using functional Magnetic Resonance Imaging (fMRI) is powerful because it enables us to directly observe how a video content stimulates the brains of individual people. We propose a brain imaging-based RNP that represents the similarity of video-evoked brain activities between people as a network graph. We show an application scenario featuring predictive content delivery using the proposed RNP in which, when a user shows interest in a video content in some way, other users close to him or her can be expected to also be interested in it because their brain activities are correlated. Through numerical evaluation using multiple real datasets obtained by fMRI, we demonstrate that the proposed RNP is generalizable across brain imaging results for different sets of video content, thus suggesting that brain imaging data can be used to robustly generate RNP for utilization as a powerful tool for estimating user preferences.
ASJC Scopus subject areas
- コンピュータ サイエンス（全般）