A Neural Network Regression Model Supported by Multi-Criteria Methods for Ranking Prediction in Sustainable Development Assessment

Machine learning models are considered to be high-potential tools for predicting problems involving multiple attributes operating on historical data. Predictive models find application in developing autonomous recommendation systems based on collected datasets. They enable the exploitation of expert knowledge to support decision-makers in different fields. This paper demonstrates the application of an artificial neural network model named multilayer perceptron (MLP) regressor to predict rankings based on past expert evaluations using multi-criteria decision analysis methods. The MLP regressor model was applied in combination with the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) method. The results of the MLP model were compared with two reference machine learning models: the Linear Regression (LR) model and the Kernel Ridge Regression (KRR) model. The prediction provided by the trained model demonstrates high consistency with the real ranking. This proves that the MLP regressor has a wide range of capabilities in developing autonomous recommendation systems that do not require the active involvement of a decision-maker. The developed methodology has been applied to predict the ranking of European countries for clean, affordable, and sustainable energy systems for society under Sustainable Development Goal 7 (SDG 7).