The Study On Detecting Abnormal Data Of Listed Companies With SM-SOM

Along with promoting economic globalization,it is more important to identify the financial information of listed companies.If we make the improper judgment about financial information,not only affects investor's judgment,are more likely to make financial information fraud.So,it is necessary to find effective ways to identify abnormal financial information of listed companies,make correct judgment according to the clustering of financial data,prevent the occurrence of financial fraud,standardize the securities market in our country,and create a good investment environment.Manufacturing industry is an important part of our country's listed companies and plays a key role in the development of the securities industry in our country.On this basis,this paper selects the manufacturing listed companies whose main business are electrical appliances as the research object,uses the comprehensive clustering method,and analyses the sample data.Not only verify the comprehensive research method has stronger practicability,also find out some cause of abnormal financial information of listed companies.At last,the article provides a method for preventing financial information distortion effectively.This paper was divided into five parts.The first part introduces the background and significance of this paper,and reviews the literature at home and abroad that are about the listed company financial abnormal.Then,we choose the research method of this article by comparing the several typical methods.At last,this paper gives the article contents and research methods,and presents the innovations and shortcomings.The second part is about some basic concepts of abnormal financial information and its forms.Then,we analyze the reasons of abnormal financial information and the influence of abnormal information.The third part is mainly introducing the research methods.First of all,we give simple mathematical definition for some variable involved in this article.Then,this paper introduces the process of the construction of the comprehensive method.It mainly involves three stages: the first stage is the preparation stage;the second phase is similarity analysis;the third stage is the SOM clustering and forecasting.Finally,the paper mainly introduces the cosine similarity and self-organizing mapping method.At the same time,we give the analytical criterion for this article.The fourth part is the empirical analysis.In this section,for evaluating the performance of this hybrid technique,we give some experiments with quarterly financial ratios of listed electrical manufacturing sector in P.R.China.Firstly,46 eligible companies are chosen from 177 electric equipment manufacturing companies,including 7 companies that appeared abnormal financial information and 39 good companies.Then,from 25 financial indicators,we get 14 financial indexes as the index analysis with the analysis of correlation.After comprehensive consideration,three listed companies that appeared abnormal financial information are chosen as samples companies.And nearly 10 years financial index data of them are analyzed,the dataset of financial ratios are divided into two parts: the normal data and abnormal data.Later,according to the five categories of financial indicators,we make more analysis about abnormal data clustering to find out which category these abnormal data fall into.In order to verify the effectiveness of the comprehensive method SM-SOM,we compare the clustering results with the results using single SOM method and the Logistic method respectively.It shows that the accuracy of the comprehensive method is higher.Furthermore,we use this method to forecast the sample data and obtain the ideal results.The fifth part is conclusion.In this part,according to the empirical analysis,we summarize the research process and results,and make the preparations for the further study on identifying the listed companies' abnormal data.