A Deep Learning Approach for Predicting Air Pollutants on the Construction Site

In recent years, the problem of air pollution has become an urgent issue causing a meaningful impact on health and environment. In urban areas, one of the main sources of pollution is air pollution on construction sites. It is characterized by multiple pollutants, among which one of the most worrying harmful substances is suspended particulate (PM2.5), causing serious damage to human health and environment. Although different monitoring systems have been recently introduced for assessing the level of air pollutants on construction sites, predicting their diffusion over time has not been explored so far, which is relevant to preserve the health of workers and people surrounding the area. To overcome this limitation, we propose a new framework based on recurrent neural networks for monitoring and predicting the spread of air pollutants on construction sites, in particular PM2.5, from known environmental conditions. The framework is composed of the following steps: (i) data preprocessing, (ii) model training, (iii) model testing, and (iv) model deployment in the construction site. Results obtained on the test set prove the reliability and usability of the proposed framework for the construction sites.