Machine learning and multi-criteria decision analysis for polyethylene air-gasification considering energy and environmental aspects

Abstract

There is a recognized need for developing renewable energies and in this regard, gasification process. As a result, much research is being conducted on gasification of different materials and particularly, plastic waste. Although several studies have investigated different types of models to predict gasification performance, there are few studies to predict polyethylene waste performance in gasification using multilayer perceptron (MLP) machine learning algorithms and interpreting them using multi-criteria decision-making methods. The main aims of this study are to develop MLP artificial neural networks and regression models to predict polyethylene gasification performance with high accuracy and to compare the accuracy of the models developed to determine the best one, moreover, to rank different tests using the TOPSIS method to determine the best and worst trials. One of the finding of this research is an MLP model developed and tuned according to hyperparameters with an accuracy of 99.93% with 0.190 of root mean square error, 0.003 of mean absolute error, and 0.058 of mean absolute percentage error. The research results represent a further step towards developing machine learning algorithms instead of classic regression models in polyethylene gasification. One regression model is needed for each output variable; however, this problem is solved by artificial neural networks by developing only one model. Moreover, the results showed that the using of TOPSIS and machine learning methods can be an efficient method for ranking and predicting polyethylene gasification properties.