基于特征选择和SSA-LSSVM的短期PM2.5浓度预测

Short-Term PM2.5 Concentration Prediction Based on Feature Selection and SSA-LSSVM

摘要: 为了提升PM_2.5浓度的预测精度，考虑到PM_2.5浓度受时间序列特征和非线性特征等原因的影响，导致了时间序列分析模型在预测PM_2.5浓度时会存在较大偏差。为此，提出一种基于特征选择和麻雀搜索算法(Sparrow Search Algorithm, SSA)优化最小二乘支持向量机(Least Squares Support Vector Machine, LSSVM)参数的定量预测模型。首先，将14个特征变量进行二进制编码，利用遗传算法结合最小二乘支持向量机对特征变量进行优选，获取最优特征子集；利用SSA算法对LSSVM的参数进行优化，构建SSA-LSSVM的PM_2.5浓度预测模型。实验结果表明，基于遗传算法进行特征选择和麻雀算法优化最小二乘支持向量机参数的模型，具有明显的预测效果。其中，该模型的RMSE和MAE分别为10.53和8.01，预测精度均高于其它模型。

Abstract: In order to improve the prediction accuracy of PM_2.5 concentration, considering the time series characteristics of the PM_2.5 concentration influencing factor dataset and the nonlinear characteristics of the data, the time series analysis model still has a large error in predicting PM_2.5 concentration. To this end, a quantitative prediction model based on feature selection and Sparrow Search Algorithm (SSA) to optimize the parameters of Least Squares Support Vector Machine (LSSVM) is proposed. First, the 14 feature variables are binary coded, and the feature variables are optimized by using the genetic algorithm combined with the Least Squares Support Vector Machine to obtain the optimal feature subset; the SSA algorithm was used to optimize the parameters of the LSSVM, and the PM_2.5 concentration prediction model of the SSA-LSSVM was constructed. The experimental results show that the model based on the genetic algorithm for feature selection and the sparrow algorithm to optimize the parameters of the Least Squares Support Vector Machine has obvious prediction effect. Among them, the RMSE and MAE of this model are 10.53 and 8.01, respectively, and the prediction accuracy is higher than other models.