How to reduce the false alarm rate beyond voting system for financial distress prediction

Financial distress prediction has increasingly become a hot topic. To enhance the predictive performance, this paper includes support vector machines, particle swarm optimization, fuzzy c-means and back propagation artificial neural network into the two-stage modelling process of business failure research. This paper use an empirical research which studies sixty-six failing corporation and sixty-six one-to-one matching non-falling corporation in Taiwan during 1971 to 2014 through utilizing existing data for the six years before bankruptcy. The developed two-stage prediction model in this research is 97.7% accurate on a validation sample. These findings protrude the efficacy of two-stage prediction models for commerce financial distress forecast and specially the importance of Support Vector Machines, Particle swarm optimization and Fuzzy c-means and coupled with back propagation artificial neural network in business failure research.