Description
TitleMobile intelligence analytics for urban smart living
Date Created2018
Other Date2018-05 (degree)
Extent1 online resource (xii, 110 p. : ill.)
DescriptionToday, as the sensing technology and mobile computing have been popularized, a variety of mobile data related to human mobility and urban geography have been accumulated in a large amount. This type of data comprehensively records the fine-grain events of our cities through “4W” aspects of information: What happened? Where it happened? When it happened? And who did it? By proper analysis, this data can be a rich source of mobile intelligence to support various location-based and real-time decision-making solutions for a broad range of urban smart living applications. Indeed, mobile intelligence analytics plays an important role in urban life because city residents often make choices under more uncertainty and can benefit more from personalized advice based on their preferences and contexts. Therefore, it is especially meaningful to develop data-driven methodologies which can effectively and efficiently guide users to make optimal decisions to achieve the goal of urban smart living. In this dissertation, we aim to address the unique challenges of urban smart living in mobile and pervasive business environments from both theoretical and practical perspectives. Specifically, we first develop a safety-aware house ranking system by considering the impact of neighborhood criminal offenses on house values. The proposed framework extracts features regarding community safety conditions of different houses, and utilizes multiply safety features to rank houses by unit value. To enhance safety-aware ranking, we introduce major characteristics of house profile to control the similarity between houses during pair-wise ranker learning. The experimental results show that the proposed method substantially outperforms the baseline learn-to-rank methods for safety-aware house ranking. Moreover, in the second study, we introduce an effective point-of-interest (POI) recommender system to consider the temporal compatibility between POI popularity and user regularity. We propose to use the massive human mobility data to profile the temporal pattern of POI popularity, and infer the regularity pattern of users based on the POI they visited through a modeling intuition ``you are where you go". We demonstrate the effectiveness of the proposed model through the extensive experiments on the real-world datasets of New York City. Finally, we introduce a zone embedding framework to identify the urban functions of city zones by studying massive origin-destination transportation data. We focus on exploiting the idea of word embedding in natural language processing domain to learn zone functions in urban computing domain by developing a novel analog from word co-occurrence to zone co-occurrence using human mobility patterns. To incorporate the contexts of human mobility in our framework, we develop the directed and temporal co-occurrence for considering mobility direction and time, and the different importance of co-occurrence for considering travel distance and zone attractiveness. The evaluation validates the proposed method and shows that the learned embeddings can comprehensively capture the urban functions of city zones. From the three studies, we conclude that mobile intelligence analytics can be powerful at disclosing patterns, relations and hidden knowledge, and it is promising to explore the power of mobile intelligence to provide location-based insights, and ultimately, to improve business performance.
NotePh.D.
NoteIncludes bibliographical references
Noteby Zijun Yao
Genretheses, ETD doctoral
Languageeng
CollectionGraduate School - Newark Electronic Theses and Dissertations
Organization NameRutgers, The State University of New Jersey
RightsThe author owns the copyright to this work.
Statistical Profile