Corporate sustainability assessment based on rough-grey set theory

Abstract

Purpose The purpose of this paper is to design an inference engine to measure the level of readiness of each bank before starting the corporate sustainability auditing process. Based on the output of the designed inference engine, the audition team can decide about the audition resources and the auditing process. Design/methodology/approach In this paper, the hybrid rough and grey set theory are used to design and create a rule model system to measure the sustainability level of banks. First, 16 rule models are extracted using rough set theory (RST), and the cross-validation of each model is done. Then, the grey clustering is used to combine the same condition attributes and improve the validity of the final model. A total of 16 new rule models are extracted based on the decreased condition attributes, and the best model is selected based on the cross-validation results. Findings By comparing the accuracy of rough-gray’s rule models and as a result of decreasing the condition attributes, a proper increase in the accuracy of all models is obtained. Finally, the Naive/Genetic/object-related reducts model with 95.6% accuracy is selected as an inference engine to measure new banks’ readiness level. Originality/value Sustainability measurement of banks based on RST is a new approach in the field of corporate sustainability. Furthermore, using the grey clustering for combining the condition attributes is a novel solution for improving the accuracy of the rule models.

Developing a risk assessment model for banking software development projects based on rough-grey set theory

“…Reducts generation has two approaches: full object reduction ( FOR ) and object-related reduction ( ORR ). ORR produces a set of reducts and decision rules through a minimal attributes subset that distinguishes on a per-object basis, while reduct with FOR creates a set of minimal attributes subset that designates functional dependencies (Karimi and Hojati, 2021).…”

“…In this phase, the validity of each model in predicting the decision attribute is examined. Among the generated models, the model with the most accuracy is selected as the inference engine of the expert system (Karimi and Hojati, 2021). The regular way to measure the validity of rules is to randomly separate the observations into a test set and a training set.…”

“…At the end of the cross-validation of each model, the max, min, average and standard deviation accuracy of iterations is given (Karimi and Hojati, 2021). ROSETTA toolkit is used to cross validate the models, and the abstract results of cross-validation of 32 rule models (8 rule models for each decision attribute) have been shown in Tables 7–10.…”

“…For any B⊆C, an equivalence relation IND (B) is defined as Equation (1), which is known as the B -indiscernibility relation. If (x , x′) ∈ IND (B), then x and x′ are indiscernible mutually by attributes from B (Karimi and Hojati, 2021). …”

“…This factor is a number between 0 and 1. If α (X) =1, the set X is considered as a crisp set and if α (X) ≺1, the set is a rough set (Karimi and Hojati, 2021). …”