BBO优化算法在时间序列预测中的应用

发布时间：2018-05-11 17:18

本文选题：生物地理学优化算法 + 时间序列预测　；参考：《兰州交通大学》2017年硕士论文

【摘要】：BBO(Biogeography-based Optimization,生物地理学优化)算法是一种新型的基于群体智能的进化算法,因其良好的全局寻优能力和鲁棒性,备受国内外众多研究者的关注,目前已广泛应用于现实生活中的优化问题中。时间序列预测与人们生活中许多实际应用息息相关,一直以来都是广大专家学者们研究的热点和难点。如何提高工程应用中时间序列的预测精度具有重要的理论价值与实际应用价值。基于ELM(Extreme Learning Machine,极限学习机)的预测模型已被广泛应用于工程应用中,并取得了良好的预测性能,ELM方法与优化算法的结合理应是提升时间序列预测精度的有利候选者。针对时间序列预测,将BBO优化算法用于ELM网络结构及其参数的优化选取,提出基于BBO算法优化ELM的BBO-ELM自适应预测方法。主要研究内容有如下几个方面:(1)研究BBO优化算法的基本理论及其数学模型,把工程应用中的优化问题转化为基于BBO优化算法的数学模型,对该模型的优化和具体实现进行深入研究,阐述BBO优化算法与其他进化算法的异同点。简述时间序列预测的基本概念及其建模方法,并在标准混沌时间序列上,对ELM方法的预测性能进行测试,测试结果表明ELM方法对非线性时间序列具有良好的预测能力。(2)针对如何选取时间序列中有效的和必需的历史信息的关键点,研究基于BBO优化算法与ELM方法结合的预测模型,优化ELM网络的输入变量选择,同时,还通过BBO优化选取ELM的隐含层节点数目及其参数(连接权值、偏置和激活函数)、正则化参数,得到BBO-ELM方法。在所提出方法的基础上,引入余弦迁移模型和混沌映射理论分别对其进行改进,得到MCBBO-ELM方法和CBBO-ELM方法。将上述方法与现有的GA-ELM等方法在同等条件下应用于Mackey-Glass混沌时间序列预测中并进行比较,实验结果显示BBO-ELM的预测性能得到明显提升,验证其有效性。(3)将所提出方法应用于网络流量预测、风电功率预测和交通流量预测实例中,实验结果表明,在同等条件下本文方法的收敛速度和预测精度优于对比方法,证实所提出方法的有效性和鲁棒性。
[Abstract]:BBO(Biogeography-based optimization (biogeographic optimization) algorithm is a new evolutionary algorithm based on swarm intelligence. Because of its good global optimization ability and robustness, many researchers at home and abroad pay close attention to it. At present, it has been widely used in real life optimization problems. Time series prediction is closely related to many practical applications in people's lives and has always been a hot and difficult point for experts and scholars. How to improve the prediction accuracy of time series in engineering applications has important theoretical value and practical application value. The prediction model based on ELM(Extreme Learning machine (extreme learning machine) has been widely used in engineering applications, and the combination of good prediction performance and optimization algorithm should be a favorable candidate to improve the prediction accuracy of time series. For time series prediction, the BBO optimization algorithm is applied to the optimization of ELM network structure and its parameters, and a BBO-ELM adaptive prediction method based on BBO algorithm to optimize ELM is proposed. The main research contents are as follows: (1) the basic theory and mathematical model of BBO optimization algorithm are studied, and the optimization problem in engineering application is transformed into a mathematical model based on BBO optimization algorithm. The optimization and implementation of the model are deeply studied, and the similarities and differences between the BBO optimization algorithm and other evolutionary algorithms are expounded. The basic concept of time series prediction and its modeling method are briefly introduced, and the prediction performance of ELM method is tested on the standard chaotic time series. The test results show that the ELM method has a good ability to predict nonlinear time series. Aiming at the key points of how to select the effective and necessary historical information in the time series, a prediction model based on the combination of BBO optimization algorithm and ELM method is studied. The input variable selection of ELM network is optimized. At the same time, the number of hidden layer nodes and their parameters (connection weight, bias and activation function, regularization parameters) of ELM are optimized by BBO, and the BBO-ELM method is obtained. On the basis of the proposed method, the cosine migration model and the chaotic mapping theory are introduced to improve them, and the MCBBO-ELM method and the CBBO-ELM method are obtained. The above methods are applied to the prediction of Mackey-Glass chaotic time series under the same conditions with the existing methods such as GA-ELM, and the experimental results show that the prediction performance of BBO-ELM is improved obviously. The proposed method is applied to network flow prediction, wind power prediction and traffic flow prediction. The experimental results show that the convergence speed and prediction accuracy of the proposed method are better than that of the contrast method under the same conditions. The effectiveness and robustness of the proposed method are verified.
【学位授予单位】：兰州交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18;O211.61

【参考文献】