Modeling of PM2.5 concentration prediction with maximum entropy method

The air quality monitor stations are sparse in many polluted cities due to its high cost, which can't provide accurate air quality for their residents. A model using Maximum-entropy (Maxent) method is built to obtain the continuous distribution of the air quality, especially PM2.5 concentration (particulate matter with aerodynamic diameters < 2.5 -m) which use the few samples from monitor stations and some environmental factors. Experiments are conducted in a somewhat polluted city Changsha located in south China to validate the proposed model. Results show that the distribution of the factory and the main road contribute much to the air pollution and the AUC value of the training and testing data is 0.957 and 0.962, far high to 0.5, which indicate the proposed model is valid. Copyright © 2016 American Scientific Publishers. All rights reserved.