Bankruptcy prediction using ELECTRE-based single-layer perceptron

For the outranking relation theory, the ELECTRE methods are one of the most extensively used outranking methods. To measure the degree of agreement and the degree of disagreement of the proposition “one alternative outranks another alternative”, the concordance and discordance relations are usually associated with the outranking relation. Instead of the traditional single-layer perceptron (SLP) developed according to the multiple-attribute utility theory, this paper contributes to develop a novel ELECTRE-based SLP for multicriteria classification problems based on the ELECTRE methods involving pairwise comparisons among patterns. A genetic-algorithm-based method is then designed to determine connection weights. A real-world data set involving bankruptcy analysis obtained from Moody's Industrial Manuals is employed to examine the classification performance of the proposed ELECTRE-based model. The results demonstrate that the proposed model performs well compared to an arsenal of well-known classification methods involving quantitative disciplines of statistics and machine learning.