Research on the Design of Financial Management Model Based on SOM-PNN Driven by Digital Economy

Abstract

In the contemporary financial landscape, institutions rely on big data to conduct comprehensive background analyses and continuous optimizations. Their primary goal is to seamlessly integrate quantitative analysis methods throughout every facet of risk management. This approach enables financial institutions to swiftly attain equilibrium in the intricate interplay between risk and income, ultimately striving for profit maximization within local and even broader domains. This study proposes a novel financial risk prediction methodology by harnessing the power of two types of artificial neural networks: self-organizing maps (SOM) and probabilistic neural networks (PNN). The amalgamation of SOM and PNN leverages their respective strengths, seamlessly integrating them into the algorithm presented in this paper. To compile and predict data, the SOM network employs a two-dimensional topological framework comprising two layers of neurons. Subsequently, the PNN model efficiently yields the final classification results by processing the output generated by the SOM model. This advanced composite model offers accelerated computation, effectively mitigates the influence of noisy data points, and significantly bolsters predictive accuracy. The effectiveness of the proposed method was demonstrated through a comprehensive financial risk analysis conducted on publicly listed companies spanning the years 2016 to 2020. The experimental results show that the SOM-PNN approach has achieved high accuracy in predicting financial difficulties within the selected company samples, surpassing an 85% accuracy rate. Even for the limited sample data, its predictive accuracy reaches 80%, outperforming alternative algorithms.