Use of artificial neural network to assess rural anthropization impacts

Abstract

This study evaluated the environmental conditions in different land occupation types in an urbanized rural area, compared their microclimates, and described their characteristics using a computational algorithm that assigned an environmental quality class for each area. The experiment was carried out in the city of Dourados-MS, Brazil, at the Federal University of Grande Dourados, between the summer of 2020 and winter of 2021. Temperature and relative air humidity data were collected to estimate temperature and humidity index (THI) during 40 days of winter (cold) and 40 days of summer (heat). The data were collected by wireless datalogger systems installed in the nine microenvironments evaluated plus INMET information. Secondly, a logical-mathematical model was developed involving an Artificial Neural Network to classify the scenarios (the environments) according to THI and human well-being index (HWBI). The proposed neural network was composed of an input layer with twelve neurons, a hidden layer with eighteen neurons, and an output layer with five neurons. The system proved to be efficient, with about 90% accuracy in its training and 80% in testing phase. As the first complex architecture built for multi-class classification of environmental comfort, the algorithm well reflected the studied environments, encompassing the interactions between natural resources and built spaces.

Expert system as a tool for management of thermal comfort in heat environments in agribusiness

“…In research conducted by Santana et al (2024), the authors stated that in a scenario of climate change, where the number of extreme high-temperature and intense rainfall events is expected to increase, strategies to integrate vegetation in densely built areas become even more important.…”