Research On Air Pollutant Concentration Hybrid Prediction Model And Its Application In Health Economic Benefit Evaluation

As the process of industrialization and urbanization continues to accelerate,the urban population continues to grow,energy consumption is increasing,and industries and motor vehicles emit a large amount of air pollutants,causing air pollution problems to become increasingly serious.High concentrations of air pollutants have a serious impact on residents’ healthy life,transportation and tourism,causing huge economic losses,and severely restricting social and economic development.Air pollution has become one of the focus issues of continuous concern around the world.In recent years,air pollutant concentration prediction models have gradually become one of the hot spots in air pollution research,and have attracted widespread attention from scholars at home and abroad.Accurate and reliable prediction of air pollutant concentrations,as one of the important bases for environmental protection departments to take measures to prevent and control air pollution,is of great importance to improve air quality levels.However,air pollutants are affected by many factors,and their data are characterized by complex nonlinearity and intermittence,which greatly increases the difficulty in the research of air pollutant concentration prediction model.Following the research idea of"construction of deterministic hybrid prediction model→construction of uncertain hybrid prediction model→application of prediction model in residents’ health economic benefit assessment",this paper applys outlier detection technology,data preprocessing technology,machine learning methods and multi-objective intelligent optimization algorithms to construct deterministic hybrid prediction model of air pollutant concentration and three uncertainty hybrid prediction models based on different strategies,enriching the research theory and methods of air pollutant concentration prediction.Then,this paper innovatively combines the prediction model with the impact of air pollutants on residents’ health and related economic benefit evaluation theories and methods,and constructs a research framework of "Application of Air Pollutant Concentration Prediction Models in the Evaluation of Residents’ Health and Economic Benefits".It provides support for calculating the impact of air pollutants on residents’ health and related economic costs,formulating air pollution prevention and control measures,and establishing a sound air quality health early warning and prevention system.The research consists of six chapters:the first chapter introduces the selected topic background,research significance,an overview of air pollution,the research framework of this paper and arrangement of the main chapters and major innovations and deficiencies of this study.Chapter 2 summarizes and concludes the current status of research on deterministic and uncertainty prediction models of air pollutant concentrations and health economic benefit assessment of air pollution.In addition,this chapter also introduces the performance evaluation indices and hypothesis testing to verify the prediction performance of the models.In Chapter 3.a deterministic hybrid prediction model of air pollutant concentration based on outlier detection,data decomposition,machine learning and multi-objective optimization algorithm is proposed,whose excellent performance is shown from multiple perspectives,including the single prediction model,outlier detection technology,data decomposition method,multi-objective optimization algorithm,and model saliency and stability test.In chapter 4,five distribution functions are introduced for data fitting and characteristic analysis,and then the multi-objective Harris Hawk optimization algorithm is proposed for performance testing and algorithm comparison.Finally,three kinds of uncertainty hybrid prediction models of air pollutant concentration based on different strategies are constructed.In Chapter 5,based on the fuzzy comprehensive assessment method,the hybrid deterministic and uncertainty prediction models,the decision-makers’ risk preferences,and the theory and method of evaluating the effects of air pollutants on residents’ health and economic benefits,this chapter implements comprehensive assessment of air quality level,conducts research on deterministic and uncertainty prediction models,and constructs a research framework of "Application of Air Pollutant Concentration Prediction Models in the Evaluation of Residents’ Health and Economic Benefits",and as a result proposes policy recommendations.Chapter 6 summarizes and generalizes the research work of this paper,and prospects the related research topics in the future.The main research contents and conclusions of this paper can be summarized as follows:First,in view of the inherent shortcomings of the single prediction model and most of the existing air pollutant concentration prediction studies focus on the improvement of prediction accuracy but ignore the enhancement of prediction stability,and seldom consider the influence of outliers on the prediction performance of the model,this paper first introduces outlier detection technologies and data decomposition algorithms for outlier detection and correction and data decomposition;Secondly,this study introduces a multi-objective optimization algorithm to optimize the parameters of the extreme learning machine model based on the prediction accuracy and prediction stability objective functions,and then builds an improved extreme learning machine prediction model;Finally,this paper selects air pollutant concentration datasets from Shenyang and Tianjin,and integrates five comparative experiments,sixteen comparative models,and the model’s significance and stability analysis results to verify the prediction performance of this proposed hybrid model.The experimental results show that:（1）Outlier detection technology and data decomposition method can effectively reduce the negative effects of outliers and data noise,and improve the prediction level of the model from the perspective of data preprocessing;（2）The hybrid prediction model using multi-objective optimization algorithms can make up for the shortcomings of a single prediction model,and enhance the prediction performance from the perspective of model improvement;（3）The deterministic hybrid model proposed in this paper is significantly better than comparison models,which can significantly improve the prediction accuracy and stability of the benchmark models.Second,in view of the previous research on air pollutant concentration prediction,which focused more on the deterministic prediction and less on the uncertainty prediction of air pollution concentration,first,this study adopts five distribution functions to fit and analyze the air pollutant concentration data,and then applys a data decomposition algorithm to decompose the air pollutant concentration data;secondly,a novel multi-objective optimization algorithm i.e.,multi-objective Harris hawks optimization algorithm is proposed,which is used to optimize the multi-objective parameters of the least squares support vector machine so as to build an improved least squares support vector machine.And then this paper constructs three different uncertainty hybrid prediction models based on a multi-objective optimization strategy,an error correction strategy and a fitting optimal distribution function strategy.Finally,air pollutant concentration datasets form Nanjing and Chongqing are selected to verify the validity of these developed models.The empirical results show that:（1）Lognormal distribution function shows stronger data fitting ability compared with other comparison functions,which lays the foundation for the subsequent uncertainty prediction research based on the fitting optimal distribution function strategy;（2）The proposed multi-objective Harris hawks optimization algorithm has excellent parameter optimization performance,which provides new references for dealing with multi-objective optimization problems;（3）The developed three uncertainty hybrid prediction models can obtain excellent interval prediction results,providing references for the uncertainty prediction research of air pollutant concentration.Third,the existing air pollutant concentration prediction research focuses on the improvement and promotion of the prediction model itself,and rarely involves the further application and expansion of the model.In order to further explore the research and application value of air pollutant concentration prediction models,first of all,this paper uses fuzzy comprehensive evaluation method to evaluate and analyze the air quality level,and determine the primary air pollutant as the research object based on the evaluation results;Then,based on the deterministic and uncertainty hybrid prediction models,the decision-makers’ risk preferences,and the air pollutant residents’ health effects and economic benefits evaluation theory and methods,this paper constructs a research framework of "Application of Air Pollutant Concentration Prediction Models in the Evaluation of Residents’ Health and Economic Benefits".As a result,this paper selects air pollutant concentration dataset from Beijing for verification,and the results show that:（1）Beijing’s air quality in recent years is showing a trend of improving year by year,but PM2.5 is still one of the most important primary air pollutants;（2）Air pollutants are easy to cause serious damage to the residents’ health,resulting in huge health economic costs;（3）The prediction model and its application research framework constructed in this paper is an in-depth expansion application research of the air pollutant concentration prediction model,which is helpful to calculate the impact of air pollutants on the health of residents and related health economic costs,and it is of great significance for the implementation of regional air pollution prevention and control measures and environmental quality management and cooperation..The main innovations of this paper are summarized as follows:First,different from the existing deterministic prediction model,this paper introduces the outlier detection technology,data decomposition method,extreme learning machine model and multi-objective optimization algorithm to detect and correct abnormal values,data decomposition and multi-objective parameter optimization,and finally proposes a deterministic hybrid prediction model for air pollutant concentration,which can further improve the prediction performance of this model;Secondly,this paper proposes a novel multi-objective optimization algorithm i.e.,multi-objective Harris Hawk optimization algorithm and the comparative experimental results of 22 multi-objective test functions and five basic multi-objective optimization algorithms show that the proposed multi-objective Harris Hawk optimization algorithm has extremely competitive optimization performance,which provides a new choice for solving multi-objective optimization problems;Third,previous air pollutant concentration prediction research focuses more on deterministic prediction and ignores the importance of uncertainty preidiction,this paper develops three kinds of air pollutant concentration uncertainty hybrid prediction models based on different strategies,which provides references for the uncertainty prediction of air pollutant concentration;Fourth,the existing air pollutant concentration prediction research focuses on the improvement of the model’s own performance,but seldom considers the further application and expansion of the model.This paper combines the deterministic and uncertainty hybrid prediction models with the evaluation theory of residents’ health effects and economic benefits,and constructs a research framework of "Application of Air Pollutant Concentration Prediction Models in the Evaluation of Residents’ Health and Economic Benefits",which helps to analyze the health effects and related economic costs caused by air pollutants on residents,and provide references for the further research and application of air pollutant concentration prediction models.The shortcomings of this study are as follows:First,this research only uses univariate time series of air pollutant concentrations without considering other influencing factors.Then,if the related factors are introduced for analysis and modeling,the prediction performance of the model may be improved,which is worthy of future in-depth study.Second,this article only evaluates the residents’ health effects and related economic benefits brought by the control of PM2.5 pollutant concentration after reaching the standard,and only selects eight health endpoints,including the premature death,respiratory diseases,cardiovascular diseases,pediatrics,internal medicine,acute bronchitis,chronic bronchitis,and asthma,without considering other air pollutants or health endpoints,as a result,the estimated values may be underestimated,which is worth further study in the future.