基于贝叶斯优化的集成模型对PM_(2.5)浓度预测

PM_(2.5) Concentration Forecasting Based on Integrated Model by Bayesian Optimization

为提升细颗粒物(PM_(2.5))浓度预测的准确性,引入大气气象参数,分析PM_(2.5)与污染物以及大气气象参数之间的相关性,并对某市62个地点使用k-means算法进行聚类,分别对聚类后的不同区域,使用集成算法建立Bagging-Stacking集成模型,对PM_(2.5)浓度进行预测。实验结果表明,使用贝叶斯优化的Bagging-Stacking模型能更准确地挖掘影响因子之间的潜在关系;与单一预测模型相比,该模型的预测结果具有更低的MAE、RMSE和更高的R^(2),表明模型具有更好的预测性能和泛化能力。

In order to improve the accuracy of PM_(2.5)concentration prediction,atmospheric meteorological parameters were introduced to analyze the correlation between PM_(2.5)and pollutants and atmospheric meteorological parameters,and k-means algorithm is used for clustering,and the Bagging-Stacking ensemble model is established for the different regions after the clustering,and the PM_(2.5)concentration is predicted by the ensemble algorithm.The experimental results show that the Bagging-Stacking model based on Bayesian optimization can more accurately mine the potential relationship between influencing factors.The model’s predictions have lower MAE and RMSE,the prediction results of the model have higher R^(2),the model has better predictive performance and generalization ability.