User interest acquisition by adding home and work related contexts on mobile big data analysis

摘要

User interest acquisition facilitates customized service by figuring out user preferences in various areas, such as recommendation system and intelligence city. Mobile Internet enriches traditional user behaviors (such as who (user), when (time) and what (content)) by introducing where (mobility) into the analysis of user interest acquisition. However, user mobility is highly predictable, and user interests are constrained in a small scope. In the era of mobile big data, although several association rules and Bayesian model based approaches have been proposed to identify user interests, the impact of home and work related contexts in users' daily lives on user interest has not been fully investigated. In fact, home and work locations are anchors in user mobility and provide abundant behavior contexts to know a person. So this article proposes a framework using home and work related contexts to identify user interests. The proposed framework consists of home-work related contexts awareness based on greedy strategy, dimensionality reduction based on principle components analysis, and modeling based on various state-of-the-art machine learning algorithms. Then the proposed framework is validated on a real dataset covering 6,800 residents with more than 3.2 million records in 23 days. Results show that the proposed framework is effective, and the precision can reach more than 82% with only 7 principle components.