Application of RBF Neural Network Based on Ant Colony Algorithm in Credit Risk Evaluation of Construction Enterprises

Abstract

To the loan offers, credit risk evaluation is the decisive link for investment. In order to evaluate credit of construction enterprises more scientifically and comprehensively, this paper establishes a systematic evaluation system, in which indexes, such as comprehensive loans status, qualities of leaders, third-party guarantee, have received due attention, and peculiar characteristics of the construction industry are full considered. As an advanced system, the Back Propagation (BP) neural network has found wide application in comprehensive evaluation, however, it increasingly shows its limitations, such as slow convergent speed and easy convergence to the local minimum points. To break through and develop, this paper proposes a new evaluation model that combined ant colony algorithm (ACA) with radial basis function (RBF) neural network, which performs better in extensive mapping ability, the evaluation accuracy, convergence rate, distributed computation of ACA and training span. Take credit status of 30 construction enterprises as samples, experimental results shows that it is effective and suitable to apply this method to credit comprehensive evaluation.