Violations detection of listed companies based on decision tree and K-nearest neighbor

摘要

Violations of listed companies to disclose accounting information will mislead the ordinary investors seriously and bring huge losses to investors. Therefore, it is particularly necessary to analyze and identify the violations of listed companies based on scientific and effective methods to avoid investment risks in advance. In this paper, we firstly use t-statistic to select eight useful and characteristic variables and build characteristic attribute space. Subsequently we construct VD (violations detection) models based on the decision tree and KNN (K-nearest neighbor) method respectively to detect violations of listed companies. The data we used come from CSMAR (China Stock Market & Accounting Research Database) and the China Securities Regulatory Commission website. The result shows the accuracy of KNN method is superior to that of the decision tree method on listed companies' violations detection.