摘要
鉴于目前鲜有研究关注P2P网贷市场收益率预测问题,针对已有金融市场收益率预测研究存在的不足,提出了一种基于两阶段分解技术和粒子群优化极限学习机的EWT-SSA-PSO-ELM预测模型.引入EWT经验小波分解算法对原始的收益率综指序列进行分解,进而提高原始序列的分解效率;采用Lempel-Ziv复杂度算法提升模式分量重构的科学性,避免以往分量重构过程的随意性;利用SSA奇异谱分解算法对高频重构分量进行降噪,从而提升高频重构分量预测效果.基于该预测模型对P2P网贷市场收益率综指进行预测,实证结果表明,所构建的收益率预测模型的性能显著优于其余基准对比模型.
Given that there are few studies focusing on the prediction of P2P online lending market yield and there are many shortcomings in the existing yield prediction studies,this paper proposes a novel hybrid EWT-SSA-PSO-ELM forecasting model based on a two-phase decomposition technique and an extreme learning machine optimized by particle swarm optimization.Firstly,this paper introduces a novel empirical wavelet transform algorithm to improve the decomposition efficiency of the original yield composite index series.Meanwhile,this paper utilizes the Lempel-Ziv complexity algorithm to enhance the reconstruction scientificity of the empirical modes,avoiding the randomness of the previous empirical modes reconstruction methods.Thirdly,this paper uses the singular spectrum analysis algorithm to perform noise reduction on the highfrequency component,further improving the prediction performance.The experimental results demonstrate that the proposed EWT-SSA-PSO-ELM model has superior prediction performance compared with each benchmark forecasting model.
作者
崔金鑫
邹辉文
Cui Jinxin;Zou Huiwen(School of Economics and Management,Fuzhou University,Fuzhou 350116,China;Institute of Investment and Risk Management,Fuzhou University,Fuzhou 350116,China)
出处
《系统工程学报》
CSCD
北大核心
2021年第3期367-381,共15页
Journal of Systems Engineering
基金
国家自然科学基金资助项目(71573042)
福建省自然科学基金资助项目(2017J01794).
关键词
P2P网贷市场收益率
EWT分解算法
SSA分解算法
PSO-ELM模型
peer to peer online lending market yield
empirical wavelet transform algorithm
singular spectrum analysis algorithm
PSO-ELM model
