Title: Studying Class Membership Scores in Machine Learning Classification for Imbalanced Binary Data
Language: English
Authors: Katzengruber, Matthias 
Qualification level: Diploma
Keywords: anomaly detection; machine learning; classification; network traffic analysis
Advisor: Zseby, Tanja 
Assisting Advisor: Iglesias Vazquez, Felix 
Issue Date: 2020
Number of Pages: 65
Qualification level: Diploma
Machine learning is getting increasing importance and is strongly promotedby the rise of computational power. A paramount application of machinelearning is anomaly detection, sometimes understood as one-class classification,i.e., a binary classification problem in which there is a significantimbalance between the minority class (anomalies/outliers) and the majorityclass (normal/inlier). Real-life cases of such scenarios are, for example, fraud detection or attack detection in network communications. In this work, we study if the assumption is correct that wrongly classified instances are closerto decision boundaries and if this information can help to refine classificationperformances. We conducted experiments on network traffic and onother imbalanced datasets and found that, as a general rule, classificationalgorithms are able to leverage class membership scores to improve the “averageprecision” metric, which is suitable for evaluating imbalanced cases.Hence, class membership scores—defined based on distances to classificationthresholds—help to improve classification while keeping the model explainabilityand the algorithm complexity simple.
URI: https://doi.org/10.34726/hss.2020.57167
DOI: 10.34726/hss.2020.57167
Library ID: AC15754408
Organisation: E389 - Telecommunications 
Publication Type: Thesis
Appears in Collections:Thesis

Show full item record

Page view(s)

checked on Feb 21, 2021


checked on Feb 21, 2021

Google ScholarTM


Items in reposiTUm are protected by copyright, with all rights reserved, unless otherwise indicated.