| 298 | 6 | 263 |
| 下载次数 | 被引频次 | 阅读次数 |
针对江苏省空气质量指数(air quality index, AQI)的预测问题,提出一个将反向(back-propagation, BP)神经网络与ε-支持向量回归机(ε-support vector regression,ε-SVR)算法相结合的SVR-BP回归算法。对训练集采用ε-SVR进行样本筛选组成新的样本集,再采用BP神经网络进行预测。样本集选取的时间跨度为2 a,样本数据为江苏省共98个监测点空气中各成分的含量。分别采用SVR-BP算法、BP神经网络和ε-SVR算法在数据更新频度不同的3个模型下对未来72 h的AQI进行预测。实验结果表明:本研究提出的SVR-BP算法的平均绝对百分误差较ε-SVR算法提升了4%～19%;训练时间比BP神经网络少0.1～2.5 s。SVR-BP算法预测AQI更为高效,在实时训练及样本筛选方面有更广阔的研究前景。
Abstract：In view of the prediction of air quality index(AQI) in Jiangsu Province, a regression algorithm(SVR-BP)is proposed, which combines back-propagation(BP) neural network with ε-support vector regression (ε-SVR). By selecting samples with ε-SVR, a new sample set can be formed, and then BP neural network is used to predict the AQI of each monitor point. The data is spanned for two years, recording the air components of 98 monitor points in Jiangsu. In this paper, three models with the different update frequencies are established to predict the AQI of the next 72 hours. Experiments show that the SVR-BP has a higher accuracy about 4%-19% than the ε-SVR algorithm, a shorter training time about 0.1-2.5 s than the BP neural network in the different model. So SVR-BP is a more efficient method to predict AQI, which has a broader research prospect in the dynamic training and the sample selection.
[1]刘华军，杜广杰.中国雾霾污染的空间关联研究[J].统计研究，2018, 35(4):3-15.
[2] ZHOU G Q, XU J M, XIE Y, et al. Numerical air quality forecasting over eastern China:an operational application of WRF-Chem[J]. Atmospheric Environment, 2017, 153:94-108.
[3]李莉，陈长虹，黄成，等.长江三角洲地区大气O3和PM10的区域污染特征模拟[J].环境科学，2008, 29(1):237-245.
[4] BATTISTA G, de LIETO VOLLARO R. Correlation between air pollution and weather data in urban areas:assessment of the city of Rome(Italy)as spatially and temporally independent regarding pollutants[J]. Atmospheric Environment,2017, 165:240-247.
[5]徐大海，朱蓉.大气平流扩散的箱格预报模型与污染潜势指数预报[J].应用气象学报，2000, 11(1):1-12.
[6]张梦瑶，黄恒君.基于改进加权马尔科夫链的兰州市空气质量预测[J].兰州财经大学学报，2018, 34(3):111-117.
[7]常恬君，过仲阳，徐丽丽.基于Prophet-随机森林优化模型的空气质量指数规模预测[J].环境污染与防治，2019,41(7):758-761.
[8]郭利进，井海明，南亚翔，等.基于卡尔曼滤波融合算法的空气质量指数预测[J].环境污染与防治，2017, 39(4):388-391.
[9] LI S C, YUAN D. Long-term power load forecast based on improved BP neural network[J]. Key Engineering Materials,2010, 439/440:848-853.
[10] LI G X, WANG Q, LI H L. Network traffic prediction based on wavelet neural network improved by genetic algorithm[J]. Information, 2013, 16(2):1047-1052.
[11] WANG K, XIE F, ZHENG C B, et al. Research on harmonic detection method based on BP neural network used in induction motor controller[C]//Proceedings of the 201712th IEEE Conference on Industrial Electronics and Applications(ICIEA), June 18-20, 2017, Siem Reap. New York:IEEE Xplore, 2017:578-582.
[12] WANG Q Z, ZHU W C, WANG B. Three-dimensional SVM with latent variable:application for detection of lung lesions in CT images[J]. Journal of Medical Systems,2015, 39(1):171.
[13] LI X Z, KONG J M. Application of GA-SVM method with parameter optimization for landslide development prediction[J]. Natural Hazards and Earth System Sciences,2014, 14(3):525-533.
[14] SCHULDT C, LAPTEV I, CAPUTO B. Recognizing human actions:a local SVM approach[C]//Proceedings of the17th International Conference on Pattern Recognition, August 26, 2004, Cambridge, UK. New York:IEEE Xplore,2004:32-36.
[15]张林，刘先珊，阴和俊.基于时间序列的支持向量机在负荷预测中的应用[J].电网技术，2004, 28(19):38-41.
[16] HABTEMICHAEL F G, CETIN M. Short-term traffic flow rate forecasting based on identifying similar traffic patterns[J]. Transportation Research Part C:Emerging Technologies, 2016, 66:61-78.
[17] CHAKARAVARTHY V T, PANDIT V, ROY S, et al. Decision trees for entity identification:approximation algorithms and hardness results[C]//Proceedings of the Twentysixth ACM SIGMOD-SIGACT-SIGART Symposium on Principles of Database Systems, June 11-13, 2007, Beijing, China. New York:ACM Press, 2007:53-62.
[18] HICKEY S J. Naive Bayes classification of public health data with greedy feature selection[J]. Communications of the IIMA, 2013, 13(2):7.
[19]周志华.机器学习[M].北京：清华大学出版社，2016.
基本信息:
DOI：10.12194/j.ntu.20191031001
中图分类号:TP18;X51
引用信息:
[1]邱敬怡,赵璇.基于SVR-BP算法的江苏省空气质量指数预测[J],2020,19(01):42-47.DOI:10.12194/j.ntu.20191031001.
基金信息:
国家自然科学基金项目(11701081,11861060);; 江苏省网络群体智能重点实验室项目(BM2017002);; 2017年度东南大学创新创业教育类教学改革研究项目(2017-cxcy-020)