Creating Financial Crises Alarm Model Base Of Data Ming

This paper analyzed the financial crisis of China Public companies, and includedthe pattern and the situation, in this paper also analyzed the causes of the financialcrisis and the approaches to prevent those disasters. Creating a effectively financialcrises alarm model is very important for the public company to face those kind ofrisks. In this paper the author evaluated different alarm model, and collected 1078public companies data as the sample from China A and B share markets. For those1078 companies, this model collected the balance statements from 2001 to 2005.According to five balance statements in five different time points of every company,the author calculated 56 accounting index, along with "audit results", there are 57indexes. The author used SQL Server 2005 Analysis Services system set up 12financial crisis alarm models by Decision Tree, Neural network, Logistic Regressionapproaches. Comparing with the current models, those model improve a lot in timing,quantity of the samples, the standard of the samples selection and data mining skills.Also in this paper the author used the SQL 2005 Integration Services integrated themodel forecast and accuracy forecast together automatically.Most of the current alarm models are based on statistical methods. The datamining methods of those models is query those valuable data from lots ofuncompleted and random data, and then reorganize those data to setup the structure ofthe model.Using SQL Server 2005 to set up those alarm models is a very good attempt tocombine the statistic functions of SQL Server 2005 and the data mining modelstogether. The developer can use different methods to analysis the data and select thebest model from the datasets in order to improve the quality of the alarm system.