Development of an algorithm for the prediction of three environmental parameters

Abstract

The concrete anticipation of environmental parameters, generated with the use of real-time data, is essential to optimize the monitoring and control of air pollution and climate change. In this paper, an algorithm based on neural networks was developed to predict controlled environmental parameters and with data acquired by IoT sensors. A comparison and analysis of the accuracy of four Machine Learning models Linear Regression, Decision Tree, Random Forest and Neural Networks is presented. The main objective of this manuscript is to predict three critical pollutants (ozone, carbon monoxide and propane) and two complementary environmental variables (temperature and humidity). The results reveal that Decision Tree and Random Forest prediction models are especially accurate for the humidity variable, while Neural Networks excel in the prediction of ozone levels. This study demonstrates the ability of advanced machine learning models to handle complex environmental data. These important findings are proposed as a starting point to be evaluated in detail in different scenarios with implications for environmental management and public health.