Text

theses

Recent years have witnessed a rapid adoption of smart mobile devices and their increased pervasiveness into people's daily life. As a result of this quick development, the demand for better mobile services is increasing with an even faster speed. Recommender systems become essential to deliver the right services to the right mobile users. For instance, Point of Interest (POI) recommendation enables us to recommend the right places to the right users based on their preferences. Also effective mobile app recommendation will help mobile users to make better utility of their smartphones. In this dissertation, we identify several unique challenges for recommendation in mobile business environments, and then introduce how we use advanced data mining techniques to address these challenges. First, rich semantic information such tags and descriptions are associated with POIs in location based services. To this end, we proposed a topic and location aware POI recommender system by exploiting associated textual and context information. Second, many mobile services are location-dependent, which means a user's choice can be influenced by location-dependent factors, in particular are the user mobility and geographical influence. User mobility refers to the factor that a user's interest would change among different regions. Geographical influence is related to the cost of the option. Therefore, it is important to capture a user's spatial choice behavior to make better recommendations. Along this line, we have proposed a geographical probabilistic factor model framework, which strategically captures user mobility and geographical influence, to model user spatial choice behavior. Extensive experiments demonstrate the effectiveness of the proposed approach. Third, services are usually organized into hierarchy structure such as category hierarchy. We then introduce a structural user choice model (SUCM) to learn fine-grained user choice patterns by exploiting hierarchy structure. Moreover, we design an efficient learning algorithm to estimate the parameters for the SUCM model. Evaluation on an app adoption data demonstrates that our approach can better capture user choice patterns and thus improve recommendation performance. Finally, privacy becomes a big issue for mobile service adoption. In particular, mobile apps could have privileges to access a user's sensitive resources (e.g., contact, message, and location). As a result, a user chooses an app not only because of its functionality, but also because it respects the user's privacy preference. We present the first systematic study on incorporating both interest-functionality interactions and users' privacy preferences to perform personalized app recommendations. Moreover, we explore the impact of different levels of privacy information on the performances of our method, which gives us insights on what resources are more likely to be treated as private by users and influence users' behaviors at selecting apps.

ETD_7424

Ph.D.

Includes bibliographical references

by Bin Liu

doi:10.7282/T3J67K7Z

ETD doctoral

The author owns the copyright to this work.