Dynamic Prediction Of Financial Distress Based On Longitudinal Data Streams

Financial distress prediction is an important area of research of financial analysis and enterprise risk management, especially dynamic financial distress prediction, as the future development trend of listed company’s management, has increasingly more to be concerned. It can predict a company’s overall financial status and business results by data flow. In the company’s life cycle, the new and effective financial distress prediction system can monitor the financial status dynamically. Though theoretic and empirical researches of financial distress prediction have made great achievements, yet there are several issues that need to be stressed or focused in the static models:(1) Sample data which is collected from a specific period is static. Thus, the demand of dynamic financial operation and relativity characteristic of financial distress prediction is not taken into account.(2) Most of existing studies in financial distress prediction do not put enough attention on financial distress prediction of an individual enterprise with longitudinal historical data. Such prediction models cannot dynamically and coherently reflect companies’financial situation in the different stages of their life cycle.Especially, recently research of financial distress prediction has become increasingly urgent. However, existing static models for financial distress prediction are not able to adapt to the situation that the sample data flows constantly with the lapse of time. Financial distress prediction with static models does not meet the demand of the dynamic nature of business operations. This paper explores the theoretical and empirical research of dynamic modeling on financial distress prediction with longitudinal data streams from the view of individual enterprise. In order to excavate financial data information deeply from vertical perspective, and predict financial distress dynamically, this research provides a new way of thinking about financial distress prediction. The main outcome of the innovative work include the followings:First of all, this paper proposes longitudinal concept drift of enterprise financial distress on the basis of financial distress theory, enterprise life cycle theory and relative financial distress definition. With the lapse of time, financial situations of an enterprise on different particular time points have different definitions, which is consistent with the law of enterprise’s development. Secondly, based on enterprise’s longitudinal data streams, dynamic prediction model for financial distress integrates financial indicator selection by using sequential floating forward selection method with dynamical evaluation of enterprise’s financial situation by using principal component analysis at each longitudinal time point. The new model dynamically predicts enterprise’s financial distress by using back-propagation neural network optimized by genetic algorithm. This model’s ex-ante prediction efficiently combines its ex-post evaluation. Finally, in empirical study, select a listed company which has a relatively long history. The company’s financial indicator data is collected from its published financial reports every six months in the period of mid-1995to mid-2010. It means that financial data of each period of January to June and each period of July to December is collected to construct the experimental data set, which contains31samples of the company. Results of dynamic financial distress prediction, compared with those of traditional static model, show that the longitudinal and dynamical model of enterprise’s financial distress prediction is more effective and feasible.