Summary of computer and Information technology doctoral thesis: Some hybrid methods in fuzzy rough set based attribute reduction in decision tables

The first is the hybrid filterwrapper algorithm proposal for finding the reduction set of the decision table using advanced fuzzy distance measurements and other measurements basing on fuzzy rough set to minimize2 the number of attributes of reduction set. Reduce and improve the accuracy of the classification model.
Nội dung trích xuất từ tài liệu:
Summary of computer and Information technology doctoral thesis: Some hybrid methods in fuzzy rough set based attribute reduction in decision tablesMINISTRY OF EDUCATION VIETNAM ACADEMY OF AND TRAINING SCIENCE AND TECHNOLOGY GRADUATE UNIVERSITY OF SCIENCE AND TECHNOLOGY ------------------------------- NGUYEN VAN THIENSOME HYBRID METHODS IN FUZZY ROUGH SET BASED ATTRIBUTE REDUCTION IN DECISION TABLES Major: Information Systems Code: 9 48 01 04 SUMMARY OF COMPUTER AND INFORMATION TECHNOLOGY DOCTORAL THESIS Ha noi, 2018 List of works of author1 Nguyen Van Thien, Nguyen Long Giang, Nguyen Nhu Son, “Attribute reduction method in decision table with attribute value domain receiving numerical value based on fuzzy rough set”, Journal on works of research, development and application of Information Technology & Communication, Journal of Science and Technology of Ministry of Information and Communications, Episode V-2, no. 16 (36), 12-2016, pp. 40-49.2 Nguyen Van Thien, Janos Demetrovics, Vu Duc Thi, Nguyen Long Giang, Nguyen Nhu Son, “A Method to Construct an Extension of Fuzzy Information Granularity Based on Fuzzy Distance”, Serdica Journal of Computing 10 (2016), Sofia, Bulgarian Academy of Sciences, No 1, 2016, pp. 13-30.3 Nguyen Long Giang, Nguyen Van Thien, Cao Chinh Nghia, “An attribute reduction method in decision table with continuous value domain based on fuzzy rough set approach”, Proceedings of 18th National Conference: Selected Problems of Information Technology & Communication - TP HCM, 05-06/11/2015.4 Nguyen Van Thien, Nguyen Nhu Son, Nguyen Long Giang, Cao Chinh Nghia, “On the Constructing of Extended Fuzzy Information Granularity based on Fuzzy Distance”, Proceedings of 18th National Conference: Selected Problems of Information Technology & Communication, Hanoi, 01- 02/10/2016, Page 371-376.5 Nguyen Long Giang, Nguyen Van Thien, Cao Chinh Nghia, “About a Direct Attribute Reduction Method on Decision Tables using Fuzzy Distance”, Proceeding of the 9th National Conference on Fundamental and Applied research of Information Technology (FAIR’9), Can Tho, 04-05/08/2016, pp. 825-835.6 Nguyen Van Thien, Nguyen Long Giang, Nguyen Nhu Son , “Fuzzy Partition Distance based Attribute Reduction in Decision Tables”, IJCRS2018: International Joint Conference on Rough Sets 2018, Quy Nhon, Viet Nam, August 20-24, 2018 (Accepted)7 Nguyen Van Thien, Nguyen Long Giang, Nguyen Nhu Son, “An Incremental Attribute Reduction Method in Decision Tables using Fuzzy Distance”, Proceedings of 21th National Conference: Selected Problems of Information Technology & Communication, Thanh Hoa, 27-28/07/2018, pp. 296- 302. INTRODUCTION Fuzzy rough set theory proposed by Dubois et al. [22, 23] is a combination of rough settheory and fuzzy set theory to approximate fuzzy sets based on a fuzzy equivalent relationdefined on the attribute value region. Since its appearance, fuzzy rough set theory is aneffective tool for solving directly attribute reduction problem on the original decision table(decision table not to through the data discretization step) to improve accuracy of theclassification model. Researches related to the attribute reduction based on approaching fuzzyrough set have been quite exciting in the past few years, including main methods such as themethod using fuzzy positive region [2, 72, 80, 92], the method using the fuzzy distinctionmatrix [34, 42, 29, 30, 69], the method using fuzzy entropy [45, 70, 71, 74, 91, 75, 33, 55], themethod using fuzzy distance [3, 8, 18]. More recently, some researchers have proposedextended methods based on different measurements defined [14, 19, 21, 30, 33, 35, 46, 47, 59,68, 85, 90, 100]. However, as with conventional methods of attribute reduction based onapproaching rough set, most of attribute reduction based on approaching fuzzy rough set arefilter methods, which means that the accuracy of the classification model is evaluated afterfinding the reduction set. The reduction set obtained only satisfies the conditions ofmeasurement preservation without having the highest classification accuracy. Therefore, thereduction set of filter methods has not optimized the number of attributes and classificationaccuracy. Nowadays, decision tables are often large and always changed, updated. The algorithmsapplication for finding reduction set based on approaching conventional rough set andextended rough set models have challenges. In case the decision tables are changed, thesealgorithms have to recalculate the reduction set on the whole decision table after changing sothat the cost of ...