Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 47183 (02)
- University: Sharif University of Technology
- Department: Mathematical Sciences
- Advisor(s): Daneshgar, Amir
- Abstract:
- One of the most heatedly debated issues in Computer Science is Outlier Detection due to its vast and substantial applications such as credit cards, Image Processing,tax fraud detection, and medical approaches. Consequently, Outlier detection has been researched within various domains and knowledge disciplines. On the other hand, the research attempts have not been sufficient to overcome this critical problem considerably inasmuch as nearly all proposed techniques are associated with a special kind of applications or datasets.Firstly, this thesis attempts to provide a precise definition which not only excludes other one’s drawbacks, but also has its distinctive merits. Three essential concepts of outlier, strong-outlier, and semi-centroid outlier are taken into consideration. Fur-thermore, several key theorems concerning the recognition of outliers which are able to detect outlierness attribute precisely and deterministically are presented. Another subtle point which deserves some words here is that the problem is debated in a corresponding graph-based version.In addition to the high ability of outlier detection, these theorems and techniques can also be exploited as training data generators in supervised techniques and as threshold checkers for unsupervised algorithms
- Keywords:
- Training Dataset ; Unsupervised Method ; Centroid ; Outliers Detection ; Subpartioning ; Semi-Centroid Outlier
Digital Object List
محتواي کتاب
Bookmark