Document Details

Document Type : Thesis 
Document Title :
MULTI-CLASSIFICATION TASKS IN IMBALANCED DATASETS: ON THE SYNYRGY BETWEEN ROBUST PAIRWISE LEARNING TECHNIQUES AND FEATURE SELECTION
لتصنيف المتعدد البيانات غير المتوازنة: تآزر التصنيف الثنائي مع اختبار الخصائص
 
Subject : Faculty of Computing and Information Technology 
Document Language : Arabic 
Abstract : Classification in imbalanced datasets is one of the recurring problems in real-world applications of classification. It considered as a challenge since it needs to deal with uneven distribution of examples in the training datasets that lead to generate sub-optimal classification models. The presence of multiple classes implies an additional difficulty since the relations between the classes tend to complicated. One class can be a minority class for some, while a majority for others. So, we proposed a Local Feature Selection Classification model using OVO (LFSC-OVO) for multi-class imbalanced datasets, to improve the performance of the classification in terms of average accuracy. LFSC-OVO is constructed based on problem decomposition and feature selection. The novelty of the proposed work resides in the level of application feature selection in the classification procedure, since feature selection has not been previously used locally for each binary problem. LFSC-OVO is validated and tested by 7 multi-class imbalanced datasets from KEEL dataset repository using different base classifiers and aggregation methods. Then, a comparative study is conducted to compare the performance of LFSC-OVO versus another method in state-of-art. LFSC-OVO shows the best performance in all scenarios of using different base classifiers and aggregation methods as a result of decreasing the effect of the majority classes on the base classifiers. 
Supervisor : Prof. Saleh M. Al-Shomrani 
Thesis Type : Master Thesis 
Publishing Year : 1440 AH
2018 AD 
Co-Supervisor : D. Aiiad A. Albeshri 
Added Date : Sunday, November 25, 2018 

Researchers

Researcher Name (Arabic)Researcher Name (English)Researcher TypeDr GradeEmail
تهاني سعد المشدقAl-Moshad, Tahani SaadResearcherMaster 

Files

File NameTypeDescription
 43834.pdf pdf 

Back To Researches Page