Air Pollution Prediction Using Machine Learning Algorithms: A Literature Review

Air pollution continues to be a problem that affects us all worldwide, since it is estimated that 7 million people die each year due to repeated exposure to pollutants that cause chronic conditions such as severe respiratory diseases, cardiovascular problems and cancer. On the other hand, climatological effects lead to the deterioration of the planet’s ecosystems. Therefore, air quality monitoring systems are the main tools used by governments to control the emission of toxic gases into the atmosphere. This makes it possible to ensure the quality of life and the general well-being of the population, as well as to strengthen the agricultural and industrial sectors. The objective of this work is to present a review of the characteristics present in the mechanisms applied in the prediction of air pollution, with the main objective of synthesizing the knowledge found, identifying the models, approaches and variables that are most studied. The results show that the hybrid models based on CNN-LTSM are the most used. Other studies use GRU and ELM that have good results when making predictions. Suspended particles (PM10 and PM2.5) are the main object of study. Multivariate models are also more accurate and efficient when it comes to forecasting.