Neural Network and Multiple Regression Models for PM2.5 Prediction in Rabigh, Saudi Arabia: A Comparative Assessment

A high concentration of fine particulate in the atmosphere has negative consequences for human health and wellbeing. Therefore, the prediction of the concentration of particles in atmospheric air is imperative so that the public is well aware of the atmospheric condition, and the standard of air quality can be properly managed. The research explores the feasibility of using neural network methods in replacement of the generally-used statistical models for prediction of the daily average concentration of PM2.5 (particulate matter having diameter & #8804; 2.5 um).24-h PMless thansubgreater than2.5less than/subgreater than observation from May 6th to June 17th, 2013, at a specific spot in Rabigh city revealed high chronological changes with an average of (36.97 & #177; 16.22 ug/m3). The results showed that the concentration surpassed the limit specified by (WHO) guideline (25 ug/m3). Nine toxic Trace Elements (TEs) that are dangerous for human health were considered in this study, including (V, S, Lu, Ni, Cl, Zn, Cu, Pb, and Cr). These trace elements were found in abundance in PMless thansubgreater than2.5less than/subgreater than (ug/m3). These trace elements were used as input and served as a basis for the formulation of NN models and (MLR) models. The research drew a contrast between the two models was found to be (2.017) - (10.596). The result showed that properly formulated and trained ANNs are effective in resolving the issues associated with for prediction cast of particulate pollution.