基于新闻舆情的信用债估值修正模型及其应用

发布时间：2018-01-07 00:20

本文关键词：基于新闻舆情的信用债估值修正模型及其应用　出处：《山西大学学报(自然科学版)》2017年01期 　论文类型：期刊论文

【摘要】：信用债估值是金融机构资产管理与风险控制的核心问题之一,主流估值方法诸如CreditGrades等模型无法捕捉违约事件新闻舆情及市场投资者情绪的变化。基于文本情感挖掘方法,把新闻舆情分为情感和语义两个维度,在CreditGrades模型基础上增加了一项信用点差,建立了量化修正的一个信用债估值改进模型。相比传统方法,新模型具有三点优势:(1)很多债券在日常市场活动中交易不够频繁,很多情况下某一段时间内市场没有交易,传统时间序列模型在对这种情况的预测会存在误差,而且新模型通过最近一段时间的文本挖掘,可以获取有关债券最新信息更加有效的预测债券价格走势。(2)传统方法过于依赖数字类的系统内推预测模型而无法规避系统风险和行业风险,新模型挖掘本文信息有效地弥补这个不足,具有较广普适性。(3)新模型具有自我进化功能,通过针对某一个金融领域不断更新模型中的情感词典、停用词词典、用户词典,模型的预测精度将会随着模型读取的文本数据的数量不断提高。实验表明,改进模型与原模型相比其估值的均方误差从0.134降为0.056,取得了明显高于传统方法的效果。
[Abstract]:The valuation of credit debt is one of the core problems of asset management and risk control in financial institutions. Mainstream valuation methods such as CreditGrades model can not capture the default events news public opinion and market investor sentiment changes. Based on text emotion mining method. News public opinion is divided into two dimensions: emotion and semantics, and a credit point difference is added on the basis of CreditGrades model. A quantitative modified credit valuation model is established. Compared with the traditional method, the new model has three advantages: 1) many bonds are not traded frequently in the daily market activities. In many cases, there is no trade in the market for a certain period of time, the traditional time series model will have errors in predicting this situation, and the new model will be mined by the text in the most recent period of time. Can obtain the latest information on bonds more effective to predict bond price trend. 2) the traditional methods rely too much on the digital class of the system prediction model and can not avoid the system risk and industry risk. New model mining this paper information effectively to make up for this deficiency, has a wider applicability. 3) the new model has the function of self-evolution, through a financial field to constantly update the emotional dictionary in the model. The prediction accuracy of stop word dictionary, user dictionary and model will improve with the number of text data read by the model. Compared with the original model, the mean square error of the modified model is reduced from 0.134 to 0.056, which is obviously higher than that of the traditional method.
【作者单位】：中国人民大学信息学院;中国银行间市场交易商协会;
【基金】：国家自然科学基金(71571183) 教育部人文社科基金(12YJA630046)
【分类号】：F832.51
【正文快照】： 债券按照发行主体类型可以分为利率债和信用债。前者主要包括国债、央行票据、政策性金融债和地方政府债券,通常被理解为政府机构信用担保,存在违约的可能性极小,收益只受利率波动的影响,可被认为无信用风险;后者又称为企业债和公司债,其风险因子除利率之外主要考虑发行主体的

【相似文献】