Model proposition for predicting sustainability classes using multicriteria decision support and artificial intelligence

Abstract

The current study proposes a novel prediction model of sustainability classes for electricity distribution companies in Brazil, based on sustainability indicators, aiming at a more effective risk management for a certain company among their competitors. Because such indicators are based on quantitative and qualitative measures and are very likely to incur imprecisions in their measures, the model to be proposed is based on a Multicriteria Decision Support, Rough Sets Theory, which allows the mathematical treatment of those imprecisions, and Artificial Intelligence, in this case, Machine Learning by rules inference. Consequently, decision tables are generated with condition attributes, sustainability indicators, and decision attributes, sustainability classes: high, medium or low. As a result, it is possible to predict sustainability classes based in temporal series of indicators and rules inference from decision tables, using RoughSets package in R and the jMAF software, demonstrating the use of five rule generation algorithms and their respective accuracies.