Supervised and Unsupervised Learning on Temporal Data Analytical Applications

PhD Thesis Proposal Defence


Title: "Supervised and Unsupervised Learning on Temporal Data Analytical 
Applications"

by

Mr. Fengchao PENG


Abstract:

Temporal data analysis has been widely applied in various areas, such as 
bio-informatics, outlier detection, and trajectory analysis. Different 
applications require a variety of machine learning methods. In this thesis 
proposal, we study the supervised and unsupervised learning methods in 
temporal data analytical applications. We first develop a time series 
classification method that is effective and efficient in monitoring 
devices that are used in wireless communication. We use active learning 
method to reduce the labeling cost when collecting training data. And we 
use Random Forest and Bootstrap methods to solve the label imbalance 
problem. Then we propose a novel active learning method for time series 
classification. The method adapts the idea of shapelet discovery and 
select the training data based on both the uncertainty of classifier and 
the utility of each data instance. Finally we study the problem of team 
strategy detection which is an important problem in team sport games. We 
propose an unsupervised method to identify trajectory patterns that match 
with frequently used team strategies.


Date:			Friay, 19 January 2018

Time:                  	4:00pm - 6:00pm

Venue:                  Room 5501
                         (lifts 25/26)

Committee Members:	Prof. Lionel Ni (Supervisor)
 			Dr. Qiong Luo (Supervisor)
 			Prof. Qian Zhang (Chairperson)
 			Prof. Lei Chen


**** ALL are Welcome ****