Air Quality Prediction Based On Integrated Algorithms

With the development of social economy and expansion of city population,air pollution has been becoming a serious problem,which has seriously affected our daily life.The State Council planed to establish monitoring and warning systems in cities of the country,which can release information and improve the accuracy of monitoring and warning system.In order to know the trend of environment pollution,and prevent serious pollution incidents,the following aspects have been studied:First of all,through the analysis of air quality in recent years in Tianjin City,find out the main pollutants and their change trend in recent years and analyzes the reasons.Secondly,improve the weight and membership degree on the basis of fuzzy mathematics,then evaluate the air quality in nearly a year the city of Tianjin.The experiments results show that the improved fuzzy evaluation system is more reasonable and accurate.Aiming at the large toxicity and huge impact on agriculture of the benzene in the air,a kind of suitable for prediction of benzene concentration in the air AdaBoost.BT integrated modeling algorithm is proposed.Firstly,build a mathematical model by using the relationship between the benzene and other ingredients in the air.Secondly,the proposed algorithm is introduced into an elastic factor on the basis of traditional AdaBoost algorithm.As the gist of training results and weights update,the elastic factor is determined by the mathematical model.Finally,during the weight updating stage,considering the time factor of samples can reduce the effect of early production data to the model.Using the air quality data of a certain city to train and verify the proposed model,experimental results show that compared with other traditional integration algorithm,this proposed AdaBoost.BT algorithm has simple operation,less prediction errors and more accurate precision(91%),which can be used in the practical production.By analyzing the characteristics of air quality and traditional integrated algorithm,an adaptive multi-dimensional AdaBoost algorithm is proposed,which is applied to predict air quality.This proposed algorithm extend traditional AdaBoost algorithm to the multidimensional with extreme learning machine(ELM)as the weak learning machine,using the grid search method to determine the key parameters of the algorithm:the number of iterations and the threshold value.According to the prediction accuracy in the process of each iteration to decide whether to update the training sample input and output weights.This algorithm,which is breaking the limits of traditional algorithms' single-step prediction,achieving the multi-step predictions.Whats more,it can fill the missing part of historical data,and the prediction accuracy is more higher.