Research Of Listing Corporation Financial Fraud Identification Model Based On Data Mining

Many listed company’s financial fraudulence is the "tumor" of capital market which detroies the publicity and fairness of capital market. However, many firms have been as the blue chip before they are disclosed to the fraudulent. Many people wonder how and what methods could be used to identify and disclose financial fraudulence. And the methods should be reliable and could be used widely. Recently computer data mining technology has been developed to be used in many fields, including financial fraudulence and so on. This paper would propose a new financial fraudulence idenfication pattern which is combined with new data mining technology based on the computer scientific’achievement.This paper first introduces the research present situation of listing corporation financial fraudulence, describes the present research of financial fraudulence recognition from the aspects of the motivation, the signal recognition and the mining methods.It clears the current research problems. Secondly, we use the CSRC’s fraud listing corporation from2000to2010and health which is the opposite for the study sample. And we established static identification model of listing corporation financial fraud and dynamic identification model of listing corporation financial fraud.Static recognition mode is mainly based on the combination of factor analysis method and the Bias classification and deals with the static index.Dynamic identification mode starts from the company’s IPO year and ends in fraud year to establish time series of feature vectors. Based on dynamic time bending (DTW) thinking, dynamic identification mode uses k-Nearest Neighbor classification algorithm to identify financial fraudulence.The results show that the static pattern recognition effect is higher than that of traditional multivariate statistical model, dynamic identification model effect is higher than that of static pattern recognition.