AVATAR: A System for the Analysis of Vast Amount of Trajectories and Roads

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering


PhD Thesis Defence

Title: "AVATAR: A System for the Analysis of Vast Amount of Trajectories and 
Roads"

By

Mr. Ye DING


Abstract

The value of large amount of trajectory data has received wide attention in 
many applications including human behavior analysis, urban transportation 
planning, and various location-based services (LBS). Nowadays, both scientific 
and industrial communities are encouraged to collect as much trajectory data as 
possible, which brings many opportunities and problems, including: 1) the raw 
data collected by the GPS device requires reorganization and preprocessing to 
enable further analysis; 2) it is expensive and challenging to store and 
process such big trajectory data efficiently; and 3) by leveraging effective 
spatial-temporal queries from the trajectory data, it enables us to discover 
various knowledge that are difficult to identify intuitively. In this thesis, 
we propose a complete system from the preprocessing of raw trajectory data, to 
the storage and query processing of the trajectories, and finally the discovery 
of the hidden knowledge from such data. For each component of the system, we 
study and solve a specific problem regarding to the component, as: 1) inferring 
the road type in crowdsourced map services; 2) exploring the use of diverse 
replicas for big location tracking data; and 3) recommending trajectories for 
effective and efficient hunting of taxi passengers.

For the road type inference problem, we propose a combined model based on 
stacked generalization to infer the types of road segments, and conduct eight 
experiments based on different classifiers to show that our method is much 
better than the baseline methods. For the use of diverse replicas for big 
trajectory data, we propose BLOT, a system abstraction which tries to find the 
optimal set of diverse replicas that suitable for the costs of a set of 
trajectory queries. Based on our greedy strategy, the experiments show that our 
solution is effective and efficient. For the taxi hunting route recommendation 
problem, we introduce the HUNTS system based on the estimations of the benefits 
of road segments, and make reasonable hunting trajectory recommendations for 
taxi drivers to make more money.

Due to the low-density, sparsity, and uncertainty of trajectories, handling 
such data may face many challenges. In this thesis, we will discuss the 
important challenges and research issues of each aspect, and compare the 
differences between our methods and the state-of-the-art.


Date:			Thursday, 12 June 2014

Time:			2:00pm - 4:00pm

Venue:			Room 3501
 			Lifts 25/26

Chairman:		Prof. John Barford (CBME)

Committee Members:	Prof. Lionel Ni (Supervisor)
 			Prof. Huamin Qu
 			Prof. Ke Yi
 			Prof. Jeff Hong (IELM)
                        Prof. Qing Li (Comp. Sci., CityU)


**** ALL are Welcome ****