Improving feature selection for classification using evolutionary computation

نوع: Type: thesis

مقطع: Segment: masters

عنوان: Title: Improving feature selection for classification using evolutionary computation

ارائه دهنده: Provider: mahsa bojari

اساتید راهنما: Supervisors: Dr. hasan khotanlou, Dr. Moharram Mansoorizadeh

اساتید مشاور: Advisory Professors:

اساتید ممتحن یا داور: Examining professors or referees: Dr. mahlagha afrasiabi, Dr. reza mohammadi

زمان و تاریخ ارائه: Time and date of presentation: 1402/11/15، 16:00

مکان ارائه: Place of presentation: Khanmohammadi Engineer Seminar, Department of Electricity

چکیده: Abstract: In many machine learning problems, there are numerous features, not all of which are essential. This is because many of them are redundant or even irrelevant, which may reduce the classification algorithm's performance. The goal of feature selection is to address this issue by choosing only a small subset of relevant features from the large original feature set. By eliminating irrelevant and redundant features, feature selection can reduce the dimensions of the data, accelerate the learning process, simplify the trained model, and improve overall efficiency. Feature selection is a challenging task due to the vast search space. Among the various methods available for feature selection, evolutionary computation methods have garnered significant attention in recent years due to their ability or potential for global search in solving feature selection problems. The aim of this research is to improve classification accuracy in feature selection problems using evolutionary algorithms. The most important challenges in the problem of feature selection include scalability, computational cost, search methods, evaluation criteria, and the number of instances. In this research, to reduce computational costs, especially on a large scale, we utilized a method that combines feature selection and sample selection, and this method served as the baseline approach in all experiments. To enhance classification accuracy, we employed a combination of the baseline method with various evolutionary algorithms as search methods. We evaluated the proposed method on 13 datasets and, after conducting various experiments, successfully achieved an improvement in classification accuracy compared to the baseline method

فایل: ّFile: Download فایل