Mining of dynamic traffic-meteorology-atmospheric pollutant association rules based on Eclat method

Abstract

With the rapid increase of urban vehicles, the atmospheric compound pollutants, notably PM${}_{2.5}$ and O${}_3$, have significantly increased and seriously affected public health. Traffic and meteorological conditions are the primary influencing factors of pollutant concentrations, and their spatial and temporal changes affect the dispersion of pollutants. Increasing use of high-resolution big data offers opportunities to explore these correlations. More extensive quantitative studies are essential for effective air pollution control. This study presents an Eclat algorithm to quantitatively reveal the relationship between traffic, meteorology and pollutants with hourly and 5-minute scale data in the urban area of Guangzhou. We establish a research framework covering temporal pollution analysis, multifactor rule mining, and spatial effects. The results show that $P{M}_{2.5}$ and O${}_3$ exhibit coordinated trends on the daily scale influenced by traffic flow and meteorology conditions, but on the hourly scale, they are negatively correlated. At the 5-minute scale, synchronized variations occur only during specific periods. This finer scale better identifies association rules for high-concentration pollutant scenarios, and non-roadside sites outperform roadside sites in mining these associations. For example, when humidity is below 37%, atmospheric pressure is 1016.2–1020.3 Pa, wind speed is 1.7–2.6 m/s, and the traffic volume on Jiefang North Road exceeds 635 vehicles every 5 min, there is a 92.86% probability that the $P{M}_{2.5}$ concentration at GYQ (a non-roadside monitoring site) will exceed 127 $\mu$g/m${}^3$. These findings enhance our understanding of how dynamic traffic and meteorological conditions impact atmospheric pollutants and provide a scientific basis for regional collaborative pollution prevention.