A pivotal element for metropolitan planning and an essential component describing the urban design is block typology, affecting the pollution concentration. Consequently, this research examines the influence of various urban block typologies on urban pollutant distribution. Four typologies are simulated by ENVI-MET software. These typologies are cubic-shaped, L-shaped, C-shaped, and linear-shaped models. Urban air quality was assessed using relative humidity, temperature, and pollution PM2.5 concentration. The performance of typologies in terms of temperature, relative humidity, and reduction of air permeability is strongly dependent on the blocks' orientation, the block shape's rotation concerning the horizontal and vertical extensions, the height of the blocks, and the type of typology. According to these parameters, the performance is different in each of these studied typologies. Regression models propose a more reliable prediction of PM2.5 when the independent variables are temperature, relative humidity, and height of buildings, among various block typologies. Hence, this article suggests a machine learning approach, and the model evaluation shows that the Polynomial Linear Regression (PLR) model is excellent for measuring air pollution and temperature.