Original scientific paper

https://doi.org/10.17559/TV-20190615044223

Index System Reduction Method Based on the Index Similarity

Ai Wang ; University of Science and Technology Beijing, No. 30 Xueyuan Road, Haidian District, Beijing, China
Xuedong Gao ; University of Science and Technology Beijing, No. 30 Xueyuan Road, Haidian District, Beijing, China

Abstract

Multi-attribute decision making (MADM) always suffers from the result inconsistency and computational complexity problem, due to numbers of redundant and relational attributes (indexes) of the initial evaluation index system. Therefore, this paper studies the index system (IS) reduction problem through selecting the most representative indicator from each index subsystem after the IS structure partition. First, we propose and demonstrate the Index Subsystem Judgement theorem to improve the efficiency of the classic system structure partition algorithm. Second, an algorithm of index system reduction based on the index similarity (ISRS) is put forward. The ISRS is able to reduce the index quantity while still keeping the index meaning. Third, we define the direction loss rate to measure the evaluation ability loss of the IS during reduction. The algorithm is tested for a synthetic dataset to compare the proposed ISRS with different index reduction algorithms, followed by an extensive experimentation with a real-world financial dataset. Experiment results illustrate that our proposed method is able to obtain more accessible and available reduction results in practice.

Keywords

direction loss rate; Index Subsystem Judgment theorem; index system reduction; index similarity

Hrčak ID:

223312

URI

https://hrcak.srce.hr/223312

Publication date:

25.7.2019.

Visits: 1.639 *