Forecasting Air Quality Index Based On Support Vector Regression

In today’s booming economy and the acceleration of urbanization and industrialization,cities have become the main gathering place for the population.Urban air quality issues are becoming increasingly prominent.Moreover,the quality of air quality has a direct or indirect impact on people’s health.Therefore,the prediction of air quality index provides a reference for people’s travel and provides a theoretical basis for related policies.This article is mainly based on support vector regression as the theoretical basis to study the issue of air quality index.First,preprocess the air quality index related data,analyze the correlation between the meteorological data and pollutant data and the air quality index;second,use the nuclear main component（KPCA）to extract the nuclear factors affecting the air quality index,and eliminate the influencing factors.Then,the innovative use of extreme-point symmetrical mode decomposition（ESMD）to stepwise decompose the historical air quality index into finite intrinsic mode functions（IMFs）and optimal global moving averages（AGM）;finally,the global artificial fish swarm algorithm（GAFSA）and adaptive weighted particle swarm optimization algorithm（AWPSO）are combined to determine the free parameters in support vector regression.The GAFSA-AWPSO-LSSVM model is constructed to predict each sequence separately,and the results are superimposed to get the final prediction result.Combining other models and error indicators illustrates the adaptability and effectiveness of the model proposed in the paper.In this paper,the air quality index,pollution factor,and meteorological data from October to February of each year in Chengdu from 2013 to 2019 are used as data samples.First,the data samples are processed for missing values and the data is normalized.Second,the investigation of the main factors affecting air quality in Chengdu showed that PM_2.5 and PM₁₀ had the highest correlation with the air quality index.Then use KPCA to extract the nuclear components of the impact factors,reduce the model input variables,reduce the model complexity,and use ESMD to decompose the air quality index,simplify the time series complexity and ensure the completeness of the data.Finally,the global artificial fish swarm（GAFSA）was first used to find the global optimal neighborhood of the free parameters in the least squares support vector regression,and then the improved particle swarm optimization algorithm was used to quickly find the optimal free parameters.The prediction results are obtained after superimposing on each sequence.By comparing single models,the ESMD-GAFSA-AWPSO-LSSVM hybrid model proposed in this paper has better prediction performance and higher accuracy.