An improved fuzzy neural network for enterprise risk assessment model design and implementation

Abstract

With the progress of technology and the development of the times, the competition among enterprises has become more and more intense, and various risks affect the development of enterprises. To improve enterprise risk assessment accuracy, the research proposes a model based on improved neural network, using the improved Latent Dirichlet Allocation (Teo-LDA) model to classify and analyze customer reviews, and combining the improved fuzzy neural network to fuzzify and analyze customer attributes and features. The experimental data showed that the accuracy of the improved Teo-LDA-based customer review analysis method was up to 0.85 and the accuracy was up to 0.82. The average accuracy of the improved fuzzy neural network model for risk assessment was 94.4%, which was 15.8% more accurate than the Back Propagation (BP) neural network model, and 5.1% more accurate than the Genetic Algorithm (GA) neural network. The results indicate that trained by adaptive hybrid method, the improved fuzzy neural network model is the most effective in achieving accurate assessment of customer churn risk in enterprises.