基于奇异谱分析和改进萤火虫算法优化BP神经网络的时间序列预测  

Time Series Prediction by Optimized BP Neural Network Based on Singular Spectrum Analysis and Improved Firefly Algorithm

作　　者：徐厦 李水艳[1] XU Xia;LI Shuiyan(School of Science,Hohai University,Nanjing 211100,China)

机构地区：[1]河海大学理学院,南京211100

出　　处：《河南科学》2021年第9期1377-1382,共6页Henan Science

基　　金：国家自然科学基金(11771120)。

摘　　要：空气质量数据与人们密切相关,故提出利用基于奇异谱分析对原始数据的去噪、网络参数的优化以及BP神经网络的系列混合预测模型,来进行时间序列预测.先采用基于时间序列去噪获得更加平稳序列来用于后续模型预测,再通过非惯性权重和变步长改进萤火虫优化算法优化网络参数,获得合理的、稳定的网络结构,然后将奇异谱分析后的去噪数据和用改进萤火虫算法优化后的BP网络用于预测,通过实际空气质量数据验证模型效果,效果良好.Air quality data is closely related to people,so it is proposed to use a series of hybrid prediction models based on singular spectrum analysis and BP neural network to perform time series prediction.Firstly,a more stable sequence is obtained based on time series denoising for subsequent model prediction,and then the network parameters are optimized by non-inertial weight and improved firefly optimization algorithm with variable step size to obtain a reasonable and stable network structure.Then,the denoised data from the singular spectrum analysis and the BP network optimized by the improved firefly algorithm are used to predict the air quality data.The model effect is verified by actual air quality data,and the effect is good.

关 键 词：奇异谱分析 萤火虫算法 BP神经网络 时间序列 

分 类 号：O29[理学—应用数学]