This research project investigates principles and methods for uncovering sophisticated patterns and actionable knowledge from massive moving object data.  Thanks to the rapid progress and broad adoption of sensor, GPS, wireless network, and other advanced technologies, moving object data have been accumulating in unprecedented scale.  However, moving object data can be dynamic, sparse, scattered, and noisy, and patterns and knowledge to be mined could be deeply hidden, sophisticated, and subtle.  This project investigates effective and scalable methods for mining various kinds of complex patterns from dynamic and noisy moving object data, finding multiple interleaved periodic patterns, and performing in-depth multidimensional analysis of moving object data.  It integrates and extends multiple disciplinary approaches derived from spatiotemporal data analysis, social media mining, text mining, statistics, and machine learning.  The study takes animal movement data, traffic data, and social media data as the major sources of data for investigation.  However, developed methods can be applied to the analysis of many other kinds of movement data for environmental study, traffic control, law enforcement, and protection of homeland security.  The study also addresses the issue of ensuring privacy and security protection while developing powerful pattern and knowledge discovery mechanisms.  Over 100 research papers, system demos, and conference tutorials have been published in various research and application forums and be integrated into the educational programs at UIUC.  Multiple graduate students were supported by this project.   The project Web site, http://www.cs.uiuc.edu/homes/hanj/projs/movemine.htm, contains more information.

Here are a few highlights on research results (several figures are included as supplements as well).

The MoveMine system has been released to animal scientists and general public:  This project generated MoveMine, a system prototype for mining interesting patterns in moving object data.  We have explored methods for sophisticated moving object data mining and developed novel techniques for mining spatiotemporal and trajectory patterns. A user-friendly interface is provided to facilitate interactive exploration of mining results and flexible tuning of the underlying methods.  MoveMine has been tested on multiple kinds of real data sets and it will benefit users to carry out versatile analysis on these kinds of data.  It will also benefit researchers to realize the importance and limitations of current techniques as well as the potential future studies in moving object data mining.  MoveMine has been used by animal scientists in MovBank.com for animal movement analysis.  A new version has been shipped to MoveBank.org in 2014.   Research publications include: (1) Zhenhui Li, Jingjing Wang, and Jiawei Han, "ePeriodicity: Mining Event Periodicity from Incomplete Observations", IEEE Transactions on Knowledge and Data Engineering (TKDE), 27(5): 1219-1232 (2015); (2)  Zhenhui Li, Jiawei Han, Bolin Ding, and Roland Kays, “Mining Periodic Behaviors of Object Movements for Animal and Biological Sustainability Studies”, Data Mining and Knowledge Discovery, 24(2):355-386, 2012; and (3)  Zhenhui Li, Jiawei Han, Ming Ji, Lu-An Tang, Yintao Yu, Bolin Ding, Jae-Gil Lee, and Roland Kays, "MoveMine: Mining Moving Object Data for Discovery of Animal Movement Patterns", ACM Transactions on Intelligent Systems and Technology (ACM TIST), 2(4):37, 2011.

Topical Trajectory Pattern Mining: Geo-coded social media data enables us to not only browse information based on locations, but also discover patterns in the location-based behaviors of users. Many techniques have been developed to find the patterns of people’s movements using GPS data, but latent topics in text messages posted with local contexts have not been utilized effectively. We develop a latent topic-based clustering algorithm, TOPTrac, to discover patterns in the trajectories of geo-tagged text messages and propose a novel probabilistic model to capture the semantic regions where people post messages with a coherent topic as well as the patterns of movement between the semantic regions.  Our experiments on real-life data sets show TOPTrac finds diverse and interesting trajectory patterns and identifies the semantic regions in a finer granularity than the traditional geographical clustering methods.  See: Younghoon Kim, Jiawei Han, Cangzhou Yuan, "TOPTRAC: Topical Trajectory Pattern Mining", Proc. 2015 ACM SIGKDD Int. Conf. on Knowledge Discovery and Data Mining (KDD'15), Sydney, Australia, Aug. 2015.

Geoburst: Real-Time Local (Anomalous) Event Detection in Geo-Tagged Tweet Streams.  We have developed an efficient method for the real-time discovery of local events (e.g., protests, crimes, disasters), which can be of great importance to various applications, such as crime monitoring, disaster alarming, and activity recommendation. This task was nearly impossible years ago due to the lack of timely and reliable data sources, but the recent explosive growth in geo-tagged tweet data brings new opportunities to it.  Our developed method has shown its high promise on both efficiency and effectiveness.  See: Chao Zhang, Guangyu Zhou, Quan Yuan, Honglei Zhuang, Yu Zheng, Lance Kaplan, Shaowen Wang, Jiawei Han, "GeoBurst: Real-time Local Event Detection in Geo-tagged Tweet Stream",   Proc. of 2016 ACM SIGIR Conf. on Research & Development in Information Retrieval (SIGIR'16), Pisa, Italy, July 2016.

Last Modified: 12/03/2016
Modified by: Jiawei Han