Uncertain data processing and applications

Abstract

Data uncertainty is inherent in many important real-world applications. In this thesis, we focus on the problem of missing value estimation in uncertain data processing. We study three applications of missing value estimation which are RFID tracking, expert finding and user tagging. We first study the problem of RFID tracking in the mobile environment. The challenging of RFID tracking is missing reading, where the read rate of RFID data in the real-world is often in the range of 60-70%. We propose a probabilistic model for RFID tracking in the mobile environment via missing value estimation. We take advantage of the spatiotemporal correlation of tracked objects for addressing the problem. We next study the problem of expert finding for question answering in the social networks. In order to provide high-quality experts, we have to obtain sufficient amount of past activities to infer the model. However, the past activities in most social network systems are rather few, and thus the user model may not be well inferred in practice. We consider the problem of expert finding from the viewpoint of missing value estimation. We then employ users’ social networks for inferring user model, and thus improve the performance of expert finding in CQA systems. We then study the problem of user tagging in the social networks. Unlike previous studies based on matrix factorization or probabilistic modelling, we propose and study the problem of user tagging from missing value estimation via collaborative tag function learning. In this learning approach, we learn the collaborative tag functions based on users’ social activities in a microblogging system and their social connections.

Date of Award	2015
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology