Prediction of Air Pollutants based on Time-Weighted Ensemble Model and Adaptive Air Quality Index Model

Air quality makes a huge difference in human health, ecological environment, economic 2 development, and global climate governance. This study introduced Time-Weighted 3 Ensemble model into the air quality prediction model and achieved good results. The 4 prediction results were consistent with reality, and the R-square of prediction is 0.54, pro- 5 viding a new reference for people to avoid air pollution. And because of the the original Air 6 Quality Index (AQI) has limited using scope and results are inaccurate, this thesis establish 7 a brandnew evaluation system, called Adaptive Air Quality Index (AAQI), which takes 8 concentration, correlation, time, and cooperation into consideration. It is more comprehen- 9 sive and advanced than the existing system. Data on six pollutants were collected from 10 six cities, namely Brasilia, Cairo, Dubai, London, New York, and Sydney, and then prepro- 11 cessed the above data using KNN interpolation, Unit transformation and normalization, 12 and calculated the correlations among them by using Mutual Information, Spearman’s 13 Rank Correlation and Kendall’s Tau Correlation. Afterwards, we incorporated it into the 14 AAQI and obtained their air quality. Among them, Sydney had the best air quality, while 15 Dubai and Cairo had relatively poor air quality. This research should be promoted and 16 applied in air quality monitoring in real life.