An Overview of Transfer Learning and Computational CyberPsychology

Abstract

Computational CyberPsychology deals with web users’ behaviors, and identifying their psychology characteristics using machine learning. Transfer learning intends to solve learning problems in target domain with different but related data distributions or features compared to the source domain, and usually the source domain has plenty of labeled data and the target domain doesn’t. In Computational CyberPsychology, psychological characteristics of web users can’t be labeled easily and cheaply, so we “borrow” labeled results of related domains by transfer learning to help us improve prediction accuracy. In this paper, we propose transfer learning for Computational CyberPsychology. We introduce Computational CyberPsychology at first, and then transfer learning, including sample selection bias and domain adaptation. We finally give a transfer learning framework for Computational CyberPsychology, and describe how it can be implemented.