时态支持向量机模型在股票操纵模式发现上的研究  

Research on Temporal Support Vector Machine Model in the Discovery of Stock Manipulation Patterns

作　　者：李博[1] 孟志青[1] 朱爱花 LI Bo;MENG Zhiqing;ZHU Aihua(School of Management,Zhejiang University of Technology,Hangzhou 310023)

机构地区：[1]浙江工业大学管理学院,杭州310023

出　　处：《系统科学与数学》2023年第2期356-378,共23页Journal of Systems Science and Mathematical Sciences

基　　金：国家自然科学基金项目(11871434)资助课题。

摘　　要：基于原始时间属性下的时态数据难以发现规律的特点,文章构建了时态支持向量机模型,该模型通过对输入时态数据的粒度变换,获得多个分类模型,从而能够发现多种规律.在此基础上,结合时态型操纵特征构建了股票操纵模式发现模型,最后在证监会披露的操纵股票真实数据上进行数值实验,实验发现细时态粒度数据的分类模型在识别一般操纵和严重操纵上效果较好,粗时态粒度数据的分类模型在识别未被操纵或轻微操纵上效果较好.在未知数据集上实验,该模型可以有效识别不同程度操纵股票的模式,其中1个时态粒度数据下添加市场差异特征的模型表现最好,识别准确率达到了98.25%.文章验证了在不同时态粒度输入下,时态支持向量机模型能够发现在原始数据上不能发现的模式特征,这对解决一些复杂规律在原始特征下难以被发现的问题具有重要借鉴意义.Based on the fact that the temporal data under the original time attribute is difficult to find regularities,this paper constructs a temporal support vector machine model,which obtains multiple classification models by transforming the granularity of the input temporal data,so as to discover multiple regularities.On this basis,combined with the temporal manipulation features,a stock manipulation pattern discovery model is constructed.Finally,numerical experiments are carried out on the real data of manipulation stocks disclosed by the China Securities Regulatory Commission.The experiments show that the classification model of fine temporal granularity data is better in identifying general manipulation and serious manipulation,and the classification model of coarse temporal granularity data is better in identifying unmanipulated or slight manipulation.Experiments on unknown data sets show that the model in can effectively find the patterns of manipulation of stocks to different degrees.Among them,the model with the addition of market differences under one temporal granularity data performs the best,with a recognition accuracy rate of 98.25%.This paper verifies that under different temporal granularity inputs,the temporal support vector machine model can find pattern features that cannot be found in the original data,which has important reference significance for solving some complex laws that are difficult to find under the original features.

关 键 词：股票操纵模式发现 时态数据 支持向量机 操纵特征 

分 类 号：F224[经济管理—国民经济] F832.51