多层线性模型与神经网络融合算法在公司债收益率预测中的应用
Application of Multi-Layer Linear Model and Neural Network Fusion Algorithm in Corporate Bond Yield Prediction
姚洪心 1黄雪妮2
作者信息
- 1. 东华大学旭日工商管理学院,上海 200051
- 2. 东华大学数学与统计学院,上海 201620
- 折叠
摘要
近年来,我国以银行间债券市场为代表的公司债持续发展,公司债券作为上市企业融资的重要渠道之一,其收益率相关的预测研究可以为市场参与者提供重要投资依据.结合两次熵权法合成的流动性代理指标,以多层线性回归与神经网络模型为基础构建公司债收益率预测模型,同时将股权市场的Fama-French三因子纳入到该融合预测模型中.研究结论表明,熵权法合成的流动性代理指标与Fama-French三因子在多个债券收益率预测模型中均显著.与线性回归或机器学习等单一模型相比,结合多层线性回归与神经网络学习的融合模型预测偏差更小且多个评价指标均为最佳,可为该问题的解决提供更科学合理的方法.
Abstract
In recent years,corporate bonds represented by the inter-bank bond market in China have been developing continuously.Corporate bonds are one of the important financing channels for listed enterprises,and the forecast research on their yield can provide important investment basis for market participants.Combined with the liquidity proxy index synthesized by the double entropy weight method,the corporate bond yield prediction model is constructed based on the multi-layer linear regression and neural network model,and the Fama-French three factors of the equity market are incorporated into the fusion prediction model.The results show that the liquidity proxy index synthesized by entropy weight method and Fama-French three factors are significant in many bond yield prediction models.Compared with a single model such as linear regression or machine learning,the fusion model combining multi-layer linear regression and neural network learning has smaller prediction deviation and multiple evaluation indicators are the best,which can provide a more scientific and reasonable method for solving this problem.
关键词
熵权法/多层线性模型/神经网络/公司债/融合算法Key words
entropy weight method/multi-layer linear model/neural network/corporate bond/fusion algorithm引用本文复制引用
出版年
2024
NETL
NSTL
万方数据