Haze Prediction Analysis Based On Support Vector Machine

With the rapid development of economy,the material and cultural life style has become more and more rich.People have begun to improve their quality of life and pay more attention to their health.Air quality has deteriorated in many places in China,and people’s normal life has gradually begun to be affected by air pollution.The term "smog",a synonym for air pollution that looks like fog but not fog,has gradually become well-known,which has caused heated debate and become a social hot spot.At this time,researchers in different fields began to analyze and study the causes of smog,the effects of smog,and the treatment of smog,and achieved a series of research results.After that,the state formulated relevant environmental pollution plans and advocated that green water and green mountains are the great innovation of Jinshan Yinshan Mountain.Through the control of pollutant emissions,afforestation and other prevention and control measures against air pollution,the harm caused by haze was alleviated to a certain extent.Although the treatment effect is significant,some areas are still affected by it,and the area still faces the problem of haze pollution.Despite the significant control effect,some areas are still affected by it,and the area still faces the problem of smog pollution.In response to this problem,this article predicts and predicts smog in real time,thereby allowing people to reduce direct exposure to smog and take effective measures to reduce damage to the body.However,smog is affected by many factors,and the accuracy and adaptability of existing smog prediction methods need to be improved.To solve this problem,this paper proposes a smog prediction model design based on SVM algorithm,which optimizes SVM algorithm by optimizing PSO algorithm,compared with LSTM prediction model with better prediction effect,and proves the feasibility of SVM prediction of smog,and PSO-SVM model is more accurate,more suitable and faster.The main contents of this paper are as follows:1.Because of the uncertainty and non-linearity of haze influencing factors,support vector machine optimization model with strong robustness is selected to establish the prediction model of haze concentration.To judge the influence of fine particles by linear analysis PM2.5 Concentration factor,analysis PM2.5 For the relationship with all kinds of influencing factors,select the appropriate influencing factors as the input of the model.2.The key parameters of SVM algorithm are adjusted by using the optimized PSO algorithm,and the PSO-SVM prediction model is established.At the same time,the LSTM haze prediction model based on time series is established as the reference experiment of the hybrid model.3.In order to evaluate the effectiveness of SVM algorithm truly and comprehensively,this paper classifies historical data as training samples and prediction samples.In the content of the experiment,PSO-SVM model is used to compare with SVM model and LSTM model.The experimental results show that the PSO-SVM hybrid prediction model has obvious advantages in prediction accuracy and adaptability,and the accuracy of prediction rate is the highest.