Research On Analysis Of Haze Weather Prediction Based On Heuristic Intelligent Optimization Algorithm And Support Vector Machine

The haze weather phenomenon occurs frequently,which has an inconvenient impact on people’s life and travel.Aiming at this problem,this paper proposes a study on forecasting and analysis of haze weather.Data prediction analysis research is to use various effective methods to estimate and judge the development trend of an event.It has been widely used and has important theoretical research and practical application value.At present,most of the research methods of data prediction and analysis mainly focus on machine learning algorithms,and support vector machine has superior prediction performance.Therefore,based on support vector machine,this paper proposes an improved method for support vector machine,and realizes the prediction of haze weather.Improvement measures are mainly proposed from the following aspects:First,propose improvements from the perspective of the kernel function and kernel parameters of support vector machines.A heuristic intelligent optimization algorithm is proposed to optimize the kernel parameters.The heuristic intelligent optimization algorithm has the incomparable advantages of other optimization algorithms,among which the firefly algorithm is the most suitable for the parameter optimization of the support vector machine.And because the firefly algorithm is prone to premature convergence phenomenon,it is proposed to improve the initial population,moving step and inert particles of the firefly algorithm to improve its global convergence ability and local search ability.A multi-kernel learning is proposed to improve the kernel function.Multi-kernel learning can maximize the performance of kernel functions and improve the accuracy and precision of data predictive analysis models.Second,propose improvements from the perspective of support vector machine predictive analysis model.It is proposed to combine the fuzzy support vector machine and the least squares support vector machine to reduce the computational complexity of the predictive analysis model and improve the influence of data noise on its accuracy,so as to avoid the over-fitting problem of the support vector machine itself.Third,propose improvement from the perspective of feature selection method of support vector machine predictive analysis model.A feature selection method based on maximum information coefficient and associated information entropy is proposed.It can better remove redundant features and reduce the impact of redundant features on predictive analysis models.An integrated feature selection method is proposed.Enhanced stability of feature subsets.The feature selection method based on maximum information coefficient and associated information entropy,the feature selection method based on maximum correlation and minimum redundancy,and the feature selection method based on search strategy are integrated to obtain the optimal feature subset for data prediction analysis.Last,establish a data prediction and analysis model integrating the improved firefly algorithm and fuzzy multi-kernel least squares support vector machine,and apply it to the field of haze weather prediction.The experimental research proves that the improved method of haze weather forecast analysis based on Heuristic Intelligent Optimization Algorithm（HIOA,Heuristic Intelligent Optimization Algorithm）and Support Vector Machine（SVM,Support Vector Machine）proposed in this paper is of practical significance.The accuracy and precision of the model are improved,the robustness of the predictive analysis model is enhanced,and the experimental results are ideal.