Comparison of artificial intelligence algorithms to estimate sustainability indicators

Abstract

The monitoring of sustainability indicators allows behavioural tendencies of a region to be controlled, so that adequate policies could be established in advance for a sustainable development. However, some data could be missed in the monitoring of these indicators, thus making the establishment of sustainability policies difficult. This paper therefore analyses the possibility to forecast the sustainability indicators of a region by using four different artificial intelligent algorithms: linear regression, multilayer perceptron, random forest, and M5P. The study area selected was the Algarve region in Portugal, and 180 monitored indicators were analysed between 2011 and 2017. The results showed that M5P is the most appropriate algorithm to estimate sustainability indicators. M5P was the algorithm obtaining the best estimations in a greater number of indicators. Nevertheless, the results showed that MP5 was not the best option for all indicators, since in some of them, the use of other algorithms obtained better results, thus reflecting the need of an individual previous study of each indicator. With these algorithms, it is possible for public bodies and institutions to evaluate the sustainable development of the region and to have reliable information to take corrective measures when needed, thus contributing to a more sustainable future.