Haze Pollution Prediction And Economic Loss Assessment Based On Ensemble Learning

With the advancement of industrialization and urbanization in China,the energy consumption is increasing,which also brings about a more serious problem of haze pollution.In this study,corresponding methods are proposed for different types of haze pollution data to predict the concentration of haze pollutants and evaluate the economic losses caused by them,with a view to providing scientific basis for haze prevention and control.For haze pollution data without index variables,such as monthly PM_2.5concentration data,this study proposes a new method called“normalized beta function gray model”to deal with the prediction problem of this type of data.Its core idea is to use normalized beta function to optimize the linear gray action quantity into a nonlinear form.When solving the nonlinear parameters in the model,the mean square error of the model is taken as the objective function to construct a constrained nonlinear optimization problem,and the simulated annealing algorithm is used to find the optimal solution under the given accuracy.And this study uses the new model to predict the monthly PM_2.5concentration in Beijing.Compared with the other three traditional models,the proposed gray model reduces the mean absolute percentage error to 32.75%,which is 60.25%lower than the maximum value of mean absolute percentage error of four gray models.For the haze pollutant concentration data set composed of univariate time series,this study proposes an ensemble model to study related issues.The main idea is to convert the single variable time series data into corresponding images,and then use the optimized Inception-v1 network to extract the implicit features from the images.Based on these features,a two-layer stacking ensemble learning framework is constructed to output the final prediction value.The new model is characterized by converting one-dimensional time series data into two-dimensional images,and automatically extracting features from the images to mine the internal relationship between data,thus improving the prediction performance of the model.The new model is used to simultaneously predict the daily average PM_2.5concentration of multiple cities.The results show that,for one-step prediction,compared with the other three classic time series prediction models,the new model reduces the mean absolute percentage error and mean absolute scaled error to 19.204%and 1.242,respectively,which are 76.607%and 77.004%lower than the maximum of them.Similarly,the new model also shows superior performance for multi-step prediction.For the haze pollution data including atmospheric factors and other indicator variables,this research improves the random forest algorithm based on the idea of ensemble learning and post-selection inference,and proposed an intelligent algorithm called“post-selection boosting random forest”to predict and analyze those data.The algorithm combines the original random forest with Lasso method.Without giving the number of decision trees for integration in advance,the algorithm can dynamically obtain the decision trees that have passed the significance test according to different input samples,and output the prediction results.The results show that compared with the random forest algorithm,the MSE value of the new algorithm in predicting the daily average PM_2.5concentration decreases by 4.93%,showing a good prediction performance.This study uses Poisson regression relative risk model,value of statistical life method,Meta analysis and other methods to evaluate the economic losses caused by PM_2.5and O₃in Beijing and Tianjin from 2016 to 2021.The results show that the economic loss caused by PM_2.5presents a fluctuating downward trend,while the economic loss caused by O₃rises first and then decreases,and the economic loss related to O₃is higher than that related to PM_2.5in the corresponding year,and its proportion in the total loss increases year by year.Therefore,it is necessary to consider the collaborative governance of PM_2.5and O₃,and reduce the economic loss caused by pollutants through scientific prevention and control of haze.