基于决策树算法的上市公司股东行为研究  

Study on the Shareholders' Behaviors in Listed Companies Based on Decision Tree Methods

作　　者：王红武[1] 朱绍涛 蔡海博 

机构地区：[1]北京工业大学经管学院,北京100124 [2]北京工业大学数理学院,北京100124

出　　处：《数理统计与管理》2017年第1期139-150,共12页Journal of Applied Statistics and Management

基　　金：国家自然科学基金项目(11372014;11072007);国家社会科学基金重点项目(13AGL003);教育部人文社科基金项目(10YJA630185);北京市自然科学基金项目(1122001)

摘　　要：本文从生态学视角出发,将研究焦点深入到上市公司内部的股东层面。通过将上市公司股东进行生态学的种群划分,刻画不同类型股东所处的生态位,研究分析不同生态位的股东群体之间的关系和行为表现。基于国泰安数据库和CCER数据库2643家上市公司股东2008-2014年四个季度约60万条记录数据,将决策树算法用于股东统计分类和行为分析上,尝试了一种具有速度快、精确度高的特点的统计分析方法;通过分析发现股东持股排名的生态位维度对股东增减持行为和程度影响最大,为进一步的治理决策提供了依据。From ecological perspective, this paper will focus into the internal shareholders of listed companies. By dividing the different populations of the shareholders in listed companies, and depicting the niches of different types of shareholder, the paper anMysis the different relationships and behaviors of shareholder populations from different niches. Based on about 600000 shareholder records of 2643 listed companies which loaded from the Guotaian Database and CCER Database, the paper applies the decision tree method in statistical classification and behavior analysis of shareholders, experiments a statistical analysis technique which results a faster and more accurate than the common methods. Through the experiments, gets the conclusion that shareholder rank in shareholder niche dimension is the most important thing that affects the reducing or increasing behavior of shareholders, builds the foundation of governance and decision accordingly.

关 键 词：统计分类 股东行为 决策树算法 大数据 

分 类 号：O212[理学—概率论与数理统计]