Research On Financial Early-warning Using Partial Least Squares BP Neural Networks

Agriculture is a strategic industry as the foundation of China.It effects the healthy development of the national economy and the stability of the society. The healthy development of agricultural listed companies is of great importance,because that agricultural listed companies play an important role as the leader of agricultural enterprises. It is necessary to establish a Financial Early-Warning system in order to prevent the financial risk of agricultural listed companies.In this case,the systerm can help the investors and creditors to aviod huge economic losses.Therefore,the Financial Early-Warning systerm has great vital significance for agricultural listed companies.Based on predecessors’ studies,we attempt to establish an accurate and helpful Financial Early-Warning systerm. In this parer,we set up the initial independent variable of Financial Early-Warning system,which is based on the financial reports’ capability of listed agricultural companies in our country,containing profitability, solvency,operating capacity,growth ability,investment profit capability and the audit report of accounting firms.In this study,we selecte43agricultural listed companies’datas from2007to2009in Shanghai and Shenzhen Stock Exchanges as the object of study.The paper are divided into six chapters:The first chapter introduce the study’s background of agricultural listed companies’Financial early-warning systerm.And in this chapter,we explain the main content,research methods and ideas.The second chapter introduces the basic theory of the study and define concepts related.The third chapter ralate to the basis of agriculture listed companies. We choose Financial-Risk-Level Index and Divisions of Sample-Companies basing on the characteristics of agriculture companies.The fourth chapter expound the basic principles and application methods of Partial Least Square-BP Neural Network. The fifth chapter use Partial Least Squares BP Neural Networks to simulate the results of validation-samples by screening, learning and training input-samples. The comprehensive correctness of simulated results reaches as high as90.2%,and expected results and practical results are basically the same. This article provide a favorable reference for the managers and investors of listed agricultural companies,to predict financial risk scientifically and accurately.The sixth chapter of the research work in this paper is summarized and the direction of further research prospect.