Past Events

PhD Defense

Data-driven Human Behavior Learning and Prediction in Smart Cities

 

Download as iCal file

Friday, September 10, 2021, 09:30am - 11:30am

 

Speaker: Zhou Qin

Location : Remote via Zoom

Committee

Prof. Desheng Zhang (advisor)

Prof. Yongfeng Zhang

Prof. Dong Deng

Dr. Ruilin Liu (external member)

Event Type: PhD Defense

Abstract: Over the last decade, enormous data are logged and collected with the development of infrastructure in smart cities, such as cellular networks and vehicular networks. Learning human behaviors by leveraging such large-scale data sets becomes extremely popular due to the accessibility to data, which further reveals insights regarding user behavior patterns. Such insights are not only rich in business value but also helpful for improving system or service performance, which benefits the users in return. For example, learning from data traffic log data in cellular networks helps enhance the resilience of cellular systems, e.g., better handling data traffic surge during crowded events, which then improves the user experience in return. Human behaviors can generally be revealed by two dimensions, i.e., offline physical mobility patterns and online data usage patterns, which can be captured by infrastructures such as cellular systems and vehicular systems. However, most existing works mainly focus on learning from a single system or approach the learning from the aggregate system level, missing the potential aid by fusing information from heterogeneous systems or integrating both aggregate system aspect and individual user aspect. In this dissertation, building upon the comprehensive data investigation on collected city-level data sets covering millions of users, we demonstrate human behaviors learning and prediction by three concrete examples with heterogeneous systems and both aggregate and individual aspects: 1) Behavior learning and prediction on traveling time: we investigate the traffic crowdedness level by designing EXIMIUS, a hybrid measurement framework with implicit sensing data from cellular signaling data and explicit sensing data from vehicular GPS data. We then show a comparative study from two sensing sources and present a data fusion-based crowdedness level predictor. 2) Behavior learning and prediction on cellular data traffic: we present a cellular data traffic prediction framework called CellPred, which combines behavioral modeling from individual user level and aggregated modeling from cell tower level. 3) Behavior learning and prediction on mobile WiFi usage: we provide a mobile WiFi usage predictor called MIMU, which dedicates to mobility regularity and access irregularity in the mobile WiFi setting. We implement and evaluate the three frameworks by real-world data collected from two Chinese cities, Shenzhen and Hefei. Our data investigation and framework evaluation results provide insights for broad human behavior learning and prediction works, which is beneficial to both the service providers and the involved users.

 

Join Zoom Meeting
https://rutgers.zoom.us/j/98870950747?pwd=TFArWjBVenc4dU1MM1dEYTJFcHU1Zz09

Join by SIP
This email address is being protected from spambots. You need JavaScript enabled to view it.

Meeting ID: 988 7095 0747
Password: 055091
One tap mobile
+16465588656,,98870950747# US (New York)
+13017158592,,98870950747# US (Washington DC)

Join By Phone
+1 646 558 8656 US (New York)
+1 301 715 8592 US (Washington DC)
+1 312 626 6799 US (Chicago)
+1 669 900 9128 US (San Jose)
+1 253 215 8782 US (Tacoma)
+1 346 248 7799 US (Houston)

Meeting ID: 988 7095 0747