基于多分类器投票表决组合的私募股权投资风险预测度研究  

Prediction on the Investment Risk of Private Equity Based on Voting Principle for Multi-Classifier Combination

作　　者：张娅萍 赵峰[2] 张杰[2] 

机构地区：[1]河南牧业经济学院会计学院,河南郑州450044 [2]山东科技大学经济管理学院,山东青岛266590

出　　处：《统计与信息论坛》2018年第1期85-91,共7页Journal of Statistics and Information

基　　金：山东省自然科学基金项目<基于数据流概念漂移的衍生金融工具风险预警与治理研究>(ZR2015GM013);全国统计科研计划重点项目<基于云计算技术的海量数据随机处理与联网交互技术研究>(2016LZ31)

摘　　要：私募股权投资基金是一种新型衍生金融投资工具,因其非公开发行与私下募集的特性,从而具有较高的投资风险。为了监测私募股权基金投资的不确定性所诱发的潜在风险,采用了多分类器投票表决组合方法构建了风险预测模型,通过改进和优化投票表决规则来提高组合分类器对不同数据类别模式的识别能力,将统计小样本测试后得到的单个分类器识别性能的先验知识作为投票表决的根据,通过投票表决时每个类别的表决阈值设置来提升组合的预测效果。在模型构建的基础上,选取了清科集团(Zero2iPo)发布的年度私募股权相关投资案例作为对象进行仿真模拟,仿真结果显示,所构建的组合预测模型具有显著的优越性,能够大幅提高私募股权投资风险预测的准确率。Private equity investment fund is a new type of financial investment derivative instruments.Because of its characteristic of non-public and private collection,private equity investment fund has a higher investment risk.In order to test potential risks of private equity fund investment uncertainty,the paper builds risk prediction model based on voting principle for multi-classifier combination.By improving and optimizing the voting rules to implement the classifier recognition ability to different data category,the paper uses priori knowledge of the single classifier recognition ability obtained after statistical small samples test as a basis for vote and enhance the prediction effect of the combination by the setting of the voting threshold of each classifier in vote.Based on the model building,the simulation experimental results of choosing annual private equity related investment case by Zero2 iPo as object show that the prediction model is better than the single classifier model,It can greatly improve the accuracy of private equity investment risk prediction.

关 键 词：私募股权 投资风险 多分类器组合 投票表决规则 

分 类 号：F830.51[经济管理—金融学]