【摘要】 为了实现城市空气质量的精准预测,针对与城市空气质量预测相关的大数据种类多、规模大、维度高和生成速度快等特点,在研究城市不同区域空气质量评价指标的基础上,提出不同区域空气质量子空间聚类分析方法,挖掘不同区域空气质量的特征。通过对不同区域进行群体划分,并利用互信息矩阵从城市功能、地形、气象条件等方面辨识与不同区域空气质量相关联的因素,构建基于随机森林算法的城市空气质量预测模型。该方法可以有效识别城市不同区域空气质量的强关联因素,避免由于关联因素的差异性对空气质量预测造成的不利影响。仿真结果表明:该方法适用于大数据的分析与处理,并具有较高的预测精度。更多还原

【Abstract】 In order to forecast city air quality accurately,taking into the related features of big data account,including numerous varieties,great scale,high-dimension and high velocity,based on city air quality evaluation indexes of different regions,the subspace clustering analysis method of different regional air quality was put forward to investigate the characteristics of different regional air quality. Through grouping of different regions,the mutual information matrix was used to identify factors related to different regional air quality from the aspects of city functions,terrain and weather conditions,etc. in order to establish the city air quality forecast model based on random forest algorithm. The method could effectively identify the strong correlation factors of city air quality in different regions,and avoid the adverse effect on air quality forecasting due to the difference of correlation factors. The simulation results showed that this method was suitable for the analysis and processing of big data,and had high prediction accuracy.更多还原