Airborne particulate matter (PM) is an amalgam of liquid droplets found in air and microscopic solid particles. The particles differ in size, shape, and chemical composition. PM has a significant impact on climate and precipitation and adversely affects human health as it can infiltrate the lungs and enter the cardiovascular system. This article explores the various PM2.5 prediction models proposed to date to predict a region's particulate matter (PM2.5) concentration. As prediction techniques evolve rapidly, this study aims to assess the various methodologies proposed for predicting PM2.5 concentration in different regions according to the factors that influence it. Various machine learning, deep learning, and statistical models have been proposed to predict hourly or daily PM2.5 concentrations in the air. The previously proposed models were compared using the RMSE, MAE, and R2 scores as the evaluation metrics. Since most of these models were region-specific and mostly used different parameters for the prediction, the comparison highlighted the need for a generalized model that could be fine-tuned based on the parameters of a particular region. Thus, this review points to the need for a high-accuracy generalized prediction model for PM2.5 that adapts to the diverse parameters region-wise.