Research On Prediction Of Enterprise's Financial Distress In The Internet Era

Research on prediction of enterprise's financial distress is an important and widely studied topic since it can have significant impact on the decisions of managers, investor, creditors, securities supervisors and the relevant state departments. The prediction model is the base and key for the quantitative research of prediction of enterprise's financial distress which includes three parts: sample data's acquisition, feature subset (predictors) optimization and the method of building the prediction model. However, little has been done for the prediction of enterprise's financial distress in the Internet era. We must try to solve the problems of which influences the Internet environment has brought to the prediction of enterprise's financial distress, what will cause the enterprise to fall into financial distress, how to build the proper prediction system for the Internet environment, how to acquire the related information from the Internet for the experimental study, how to choose the right method to build the prediction model for the dynamic data in the Internet environment, and so on. By a method of"combining theoretical study and experimental study, combining quantitative analysis and comparative analysis", this dissertation will make an intensive study on the prediction model of enterprise's financial distress in the Internet environment, and it will be helpful to enrich the study of prediction of enterprise's financial distress in China.Based on the past studies, considering the reality of our country and the necessity of data acquisition for the experimental study, it uses listed enterprises as sample enterprises so that being specially treated is the classified standard. Then sample enterprises are divided into two groups: normal enterprises and crisis enterprises, and prediction of enterprise's financial distress becomes a two-class classification problem of enterprise's future business status. Surrounding the above problems, the main contents are as follows:(1) The dissertation discusses the impacts of Internet environment on the prediction of enterprise's financial distress. It analyzes the definition and characteristics of Internet environment. Then from three aspects of predictors, sample data and the prediction method, it analyzes the impacts of Internet environment on prediction of enterprise's financial distress.The analyses show that the information technologies have great influence on prediction of enterprise's financial distress. Considering the new causes to financial distress in the Internet environment, we'll build a comprehensive predictor system, we must make full use of information technologies to help us acquire information from the Internet, build the prediction model of enterprise's financial distress by data mining technologies, and help the enterprise avoid falling into financial distress.(2) The dissertation systematicly analyzes the causes for enterprise's financial distress in the Internet environment. From the inner and outer environment of the enterprise, it analyzes the causes generally. The inner factors come from the process of production, management and sale which includes investment risk, technology risk and the morality of entrepreneur. The outer factors come from the environment which includes economic cycle, national policy and management system, the influence of WTO, marketing risk, guarantee risk and credit insurance risk. Then it anatomizes the new causes in the Internet environment.The analyses show that economy globalization, more closely-cooperative supply chain, enterprise's digitalization degree, Internet safe risk and intelligent capital have great influence on the enterprise's financial distress in the Internet environment. Analyzing the new causes will help enterprise build a scientific predictor system for the Internet environment.(3) The dissertation discusses how to build a predictor system for prediction model of enterprise's financial distress in the Internet environment. First it analyzes the principles of building the predictor system. Then based on the principle, considering the new causes in the Internet environment, it builds a scientific predictor system especially for the Internet environment.The analyses show that we construct a predictor system for the Internet environment which includes financial ratio predictors and non-financial predictors. Financial ratio predictors come from the financial statements which are divided into six categories: capital composition, debt payment, activity, cash flow, profitability and growth. Outside the financial statements, we get non-financial predictors which reflect the economic environment, industry information, enterprise's characteristics and adaptive capability for the Internet environment. Adaptive capability predictors reflect the new causes for financial distress in the Internet environment which include intangible assets, supply chain and economy globalization predictor. The predictor system tries to reflect the enterprise's business status completely and truly so that the prediction model can evaluate the enterprise's status correctly.(4) The dissertation discusses web resources'data mining (WRDM) and its application in the prediction of enterprise's financial distress. It elaborates WRDM's two main processes: data acquisition and knowledge creation. Data acquisition has two phases: web information retrieval and web information extraction. Web information retrieval is to collect information from the Internet by three methods of manual gather, automated crawling or data query. Web information extraction is to extract the specific information. Knowledge creation is to create knowledge that represents the significant patterns in the data. The major technologies used for knowledge discovery are reporting/OLAP, pattern discovery and relevance ranking. Then it discusses WRDM's application in the prediction of enterprise's financial distress.The analyses show that WRDM can be used to help the prediction of enterprise's financial distress. We design a program of web resources'acquisition by which we acquire sample data from the Internet, use SVM to mine the sample data and seek the internal correlation between symptom information and enterprise's status, and obtain the classification function. By the function we can predict the enterprise's future status using the current related data and help the enterprise avoid falling into financial distress.(5) The dissertation discusses the methods of building the prediction model of enterprise's financial distress in the Internet environment and does some experimental studies. Considering the volatile Internet environment, it uses artificial intelligent expert system technology to build the prediction model which has function of machine-learning and can handle dynamic data effectively. It innovatively proposes a new prediction model of PSO-SVM which integrates PSO with SVM. The PSO-SVM model uses PSO to optimize both the feature subset and parameters of SVM simultaneously to improve its performance. Then it does some experimental studies by MDA, Logit, BP neural network, SVM and PSO-SVM model.The analyses show that we obtain the near optimal parameters (C=90.52,δ²=12.93) and feature subset (25FS) which includes financial ratio predictors and non-financial predictors. No-financial predictors include predictor of x₃ (the ratio of intangible assets to total assets) which reflects the enterprise's adaptive capability for the Internet environment. It shows intangible assets become the focus of financial management with the development of knowledge economy and have great impact on the enterprise's development. After comparing the results of MDA, Logit, BP neural network, SVM and PSO-SVM models, we find that PSO-SVM model shows a better result than the other models with the accuracy of 90.30%. It testifies the validity and feasibility of PSO-SVM model.The dissertation has the innovations as follows:(1) Considering the analyses of new causes for enterprise's financial distress in the Internet environment, the dissertation constructs a predictor system for the Internet environment which includes financial ratio predictors and non-financial predictors. It adds several new predictors reflecting the new causes in the Internet environment: intangile assets, supply chain and economy globalization predicator which make the predicator system reflect enterprise's business status completely and truly so that the prediction model can have an accurate evaluation for enterprise's business status.(2) The dissertation discusses the way that acquires data from the Internet—web resources'data mining and its application in the prediction of enterprise's financial distress. With the high development of network and information technology, web resources become a big potential information database including variety of information related with the enterprise. It designs a program of web resources'acquisition by which we acquire multi-snippet data from two authorized websites of listed enterprises (Shang Hai stock exchange website and Shen Zhen stock exchange website) to form the multi-dimension sample data for experimental study.(3) Both the past studies and the dissertation's experimental study on SVM model show that feature subset selection and parameters optimization have a great influence on the performance of SVM model. To improve SVM model's performance, the paper innovatively proposes a PSO-SVM model which integrates PSO with SVM. PSO-SVM model takes the wrapper approach in which feature subset selection is dependent on the learning algorithm used to construct the classifier. PSO-SVM model uses SVM as the classifier, feature set and parameters (C,δ²) of SVM as the position of particle in PSO and the classification result of SVM as the fitness value for the particle. It uses PSO to optimize both the feature subset and parameters simultaneously. After eliminating the irrespective and redundant feature, when it obtains the near optimal feature subset and SVM parameters, it also obtains the near optimal prediction result.