基于集成学习的二次协同数据预测及优化方法  

Quadratic Collaborative Data Prediction and Optimization Method Based on Ensemble Learning

作　　者：梁丽娜 张宇 张嘉玮 LIANG Li'na;ZHANG Yu;ZHANG Jiawei(Beijing Institute of Tracking and Telecommunications Technology,Beijing 100094,China;Shandong Judicial Police Vocational College,Ji'nan 250014,China;College of Computer and Cyber Security,Hebei Normal University,Shijiazhuang 050024,China)

机构地区：[1]北京跟踪与通信技术研究所,北京100094 [2]山东司法警官职业学院,山东济南250014 [3]河北师范大学计算机与网络空间安全学院,河北石家庄050024

出　　处：《现代信息科技》2025年第7期29-39,46,共12页Modern Information Technology

摘　　要：常用的空气质量预测模型对未知情况的预报效果欠佳,实际气象条件对空气污染物浓度影响显著。为降低气象条件给模型预报污染浓度带来的误差,获取预报准确性良好的模型意义重大。为此,文章提出一种基于集成学习的二次协同数据预测及优化方法。首先,把实测数据与一次预测数据相结合,针对缺失及偏离正常分布的数据,运用Fancyimpute库进行数据插补;其次,借助集成学习中的BaggingRegressor模型构建二次模型,从整体到个体剖析气象条件对污染物浓度的影响程度,通过投票机制综合所有预测结果,得出集成预测结果;最后,构建协同数据预测模型,纳入位置关系和风向因素进行综合预测。实验结果显示,该方法能有效提升数据的预测准确性,且协同预测模型提高了监测点的预测精度。The commonly used air quality prediction model has poor prediction effect on unknown conditions,and the actual meteorological conditions have a significant impact on the concentration of air pollutants.In order to reduce the error caused by meteorological conditions to the model prediction of pollution concentration,it is of great significance to obtain a model with good prediction accuracy.Therefore,this paper proposes a quadratic collaborative data prediction and optimization method based on Ensemble Learning.Firstly,it combines the measured data with primary predicted data,and uses the Fancyimpute library for data interpolation for missing and deviating from the normal distribution data.Secondly,the BaggingRegressor model in Ensemble Learning is used to construct a quadratic model,and the influence of meteorological conditions on pollutant concentration is analyzed from the whole to the individual.The voting mechanism is used to synthesize all the prediction results,and the ensemble prediction results are obtained.Finally,a collaborative data prediction model is constructed,and the location relationship and wind direction factors are included for comprehensive prediction.The experimental results show that the method can effectively improve the prediction accuracy of the data,and the collaborative prediction model improves the prediction accuracy of the monitoring points.

关 键 词：Fancyimpute库 数据插补 集成学习 BaggingRegressor模型 二次模型 协同预测模型 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]