The Application Of Multidimensional Single Index Model In Predicting Financial Distress Based On Period Gene

As the economic globalization goes deeper, international competition amongenterprises is becoming much more intensified. Meanwhile affected by the unstablemarket environment, the internal environment in which enterprises can exist anddevelop is full of uncertainty. For managers of the listed companies nowadays,prediction of enterprise financial distress can be beneficial for the companies’management and decision making.Generally speaking, all present prediction models of enterprise financial distressmainly include the original single variable model and the most recently usedmultivariate model (Logit regression analysis model, neural network model, SupportVector Machine, etc.) The majority methods would take the data of previous year ofenterprise financial crisis as the study object, and select financial index as modelvariable to build a static prediction model. However, this kind of static model wouldsurely fail to take different factors which change with time into consideration andcannot effectively apply the operation accumulated by the enterprise into the modelbuilding.The occurrence of enterprise financial distress is not only influenced by theenterprise’s management and performance of the previous year, but also effected bythe cumulative impact of time factor. Therefore, to take the function of time and indexinto account is a comparatively scientific means of building the enterprise financialdistress prediction model. Based on the theory of panel data and period gene, as wellas combined factors of time and index, this thesis managed to build a dynamicprediction model. Referred to four years’ data before the enterprise financial distressyear, as the same time, introduced the related index of corporate governance, thethesis has made a data protocol by the discrete method and enhanced the datadimension of the model. Meanwhile in the discretization process of financial indexand corporate governance index, to reflect the variation tendency of the data frommultiple perspectives, the thesis has adopted the method of multidimensionaldiscretization to process index data from two dimensions of sub-sectors discretizationand year-on-year discretization. The dynamic index selection helps to generate thebest mode by choosing reasonable index combination and then establish the best mode of ST company and the best mode of non-ST company. At last, empiricalevidence shows the single index multidimensional discretization based on the periodgene has achieved a good effect in the financial distress prediction.