Application of the Poisson process model for the early detection of enterprises' bankruptcy (Q2711714)

If banks want to maximize their profits then they want for that purpose to grant loans to reliable clients. When those clients are enterprises, the banks decide they are reliable on the basis of their economic situation. So the early detection of enterprises' bankruptcy is a crucial problem for banks. The classical methods to identify whether a company will be in bankruptcy or not are based on its accounting. The financial diagnostics rely on a fundamental technique of financial analysis; the ratios between key entries of the accounting. The art of the financial analyst is to use significant and informative ratios only.NEWLINENEWLINENEWLINEThis paper describes the construction of a tool with a view to selecting significant variables to detect enterprises' bankruptcy. Based on a Poisson process model, it uses a nonparametric step forward procedure to estimate densities and produce a discriminant analysis method. An important part of this work is the adaptation of kernel density estimation to the structure of the enterprises' data. The results of comparison experiments with classical methods (linear and quadratic as well as the \(k\)-nearest neighbour) ~are also described.