基于HULM的可转换债券和股票收益率研究

A Research on the Return of Convertible Bond and Stock Market:Based on HULM Model

首先对我国可转债市场与A股之间的相关关系进行了检验.结果显示可转债市场指数先于股票市场指数.其次使用HULM（Hidden-Unit Logistic Model）对可转换债券指数的时间序列数据进行了分类拟合和预测.通过在每个时间节点设置多个随机隐藏单元,算法能够很好地刻画时间序列中的隐藏结构并能以较高的准确度对可转换债券指数的走势进行拟合,这将为可转债收益率和股市的收益率研究提供一种全新的视角.

This article examine the interaction between Chinese stock and convertible bond market. The corresponding result shows that the convertible bond index leads the stock ones.Then we use Hidden-Unit Logistic Model （HULM） to classify the time series data of convertible bond market. By setting up multiple hidden units in every time spot, HULM can depict the latent structure of time series data. The result suggests that HULM can fit and forecast the convertible bond index through classification with high accuracy. We believe this would provide a new direction to the study of convertible bond and stock market.