Development of an Air Quality Forecasting Model Based on Neural Networks and Machine Learning Methods

The paper presents the main results on the design and development of a model for predicting the state of air quality based on the parameters: PM2.5, PM10, CO
2
, NO2, SO
2
. A methodology for selecting the optimal model for each of the predicted parameters has been developed and comparative characteristics and methods are presented. The models Prophet, Sarima, LightGBM are used as a basis. The results allow selecting the optimal model for each of the parameters, which allows you to increase the accuracy for newly added parameters by selecting the optimal model. The main results on training and data preprocessing are also presented, allowing one to estimate the time and performance costs for various types of data.