The Research On Financial Crisis Prediction Of Corporations With Ellipse Probabilistic Neural Network Ensemble

In the economic globalization process, due to the rapid development of the capitalmarkets, business competition has become increasingly fierce. Meanwhile, with theunbalanced understanding of their own internal, the risk of being deep in crisis isgrowing. Therefore, the prediction of corporate financial crisis has important theoreticaland practical significance.Based on the analysis of corporate financial risk, according to the selectingprinciple of evaluation index system of enterprise financial crisis, then this articleconstruct an evaluation system contains20indicators, through factor analysis eventuallyit forming five common factors with economic implications indicated more clear, andthus as the inputs of corporate financial distress prediction model. Meanwhile,according to the common factor scores, the paper derived composite score for eachsample, took whether a company ST as auxiliary reference. We used this model dividethe three financial condition score range, which corresponds to the three financialsituations: health, concern, crisis, abandoned the criteria simply have ST and non-ST forthe classification.Currently, using neural networks to predict enterprise financial crisis attracted manyresearchers attention. In view of this, this paper introduces the principle of probabilisticneural network and Adaboost integrated algorithm. First, improve the existing neuralnetwork to obtain oval probabilistic neural network. Then employ any combination ofweak classifiers to improve for the strong classifier Adaboost algorithm to integrate theneural network to build prediction model of enterprise financial crisis. This methodconsiders predict outcomes of each sub-neural network classifier to predict the degreeon the final results, which provides a new idea for the relevant study.Finally, we need do an empirical research, the paper selected200listed companieswhich are the cornerstone of China’s capital market as the samples, in order to increasethe practicality of the model, and this paper applied the enterprise annual financialreport data in year (t-2) to predict the financial situation in year t. We compared themodel predictions, showing that the model of probabilistic neural network ensemblebased on elliptic gets relatively high prediction for the accuracy of enterprise financialcrisis, which proved that the method is effective and feasible. So we know this researchhas certain guiding significance prediction and application value for the financial crisison China’s enterprises.