2. An evolutionary modularized data mining mechanism for financial distress forecasts

Abstract

More precise forecasting of corporate financial distress provides important judgment principles to decision-makers, such as bank loan officers, creditors, stockholders, bondholders, government officials and even general public. In this article, we introduce a modularized financial distress forecasting mechanism based on evolutionary algorithm, which allows using any evolutionary algorithm, such as Particle Swarm Optimization, Genetic Algorithm and etc., to extract the essential financial patterns. One more evaluation function modules, such as Logistic Regression, Discriminant Analysis, Neural Network, are integrated to obtain better forecasting accuracy by assigning distinct weights, respectively. For eliminating unreasonable results among these modules, the rule-based evaluation criteria are designed in our mechanism. From our experiments, applying evolutionary algorithm to select critical financial ratios obtains better forecasting accuracy, and, a much better accuracy is obtained if more function modules are integrated in our mechanism.