Abstract:
Air pollution has turned to a critical environmental problem nowadays. Prediction of air quality then plays a significant role in notifying or warning people about and controlling air pollution in every countries including Thailand. Based on data measured by the eight monitoring stations located in Rayong, Chon Buri and Chachoengsao, the simple design, Multi-Layer Perceptron or MLP, was built for neural network models to appraise and predict the air quality index or AQI in the eastern area of Thailand. The study results indicate that O3 and PM10 respectively play the dominant role in AQI value while NO2, SO2 and CO together account for less than 2% importance. The period associated with AQI levels is classified into three groups. The low AQI is at the end of summer and in rainy season (April, June to September). The medium AQI is in summer and at the beginning of rainy season (February, March and May). The high AQI is in winter (October to January). Additionally, the obtained neural network models are able to rather perfectly predict and classify the AQI groups, as seeing of the accuracy of high percentage for correct classification rate or CCR with approximately 90% in training data set as well 88% in validation data set.