Publication Date
2017
Document Type
Thesis
Committee Members
Bin Wang (Committee Member), Zhiqiang Wu (Advisor), Yan Zhuang (Committee Member)
Degree Name
Master of Science in Electrical Engineering (MSEE)
Abstract
The problem of outlier detection has received increasing attention recently because it plays a great role in many fields such as credit fraud detection, cyber security, etc. Machine Learning approach is an excellent choice for outlier detection due to its accuracy and efficiency. Outlier detection problem is unique due to the so-called classes imbalance: the inliers are extreme majority and the outliers are minority. Ensemble methods are popular in classification and regression task in practice to improve the performance of machine learning algorithms. Bagging and boosting are two common methods of them. In this thesis, we want to show the performance of bagging and boosting compared with base algorithms in outlier detection. First of all, some basic algorithms for outlier detection are described for both supervised and unsupervised methods. Next, theoretical analysis and strategies of ensemble are discussed. Furthermore, groups of experiments are conducted and the experiment results confirm the effectiveness of bagging and boosting methods for outlier detection problem.
Page Count
53
Department or Program
Department of Electrical Engineering
Year Degree Awarded
2017
Copyright
Copyright 2017, all rights reserved. My ETD will be available under the "Fair Use" terms of copyright law.
