PNN-SOM as an alternative to PNN to enhance the performance of large bankruptcy datasets
将监督学习的概率神经网络与无监督的自组织映射结合，提出PNN-SOM模型，在大型破产数据集上实现高预测精度和低计算时间，为银行和金融部门提供信用评估工具。
Among its many successful applications, probabilistic neural network (PNN) has also proved to be extremely useful in bankruptcy prediction. In our study, we have combined PNN, a supervised learning model with Self-Organising Map (SOM), an unsupervised one; aimed at developing a new model capable of efficiently predicting bankruptcy through an optimal balance between predictive accuracy and computational efficiency. The resultant, PNN-SOM has emerged as a promising solution delivering comparatively high accurate results in significantly less processing time. Various kernels like Gaussian, Triangular, Epanechnikov, Uniform, Laplacian, RBF and Exponential Kernels have been applied in this study. The robustness of our PNN-SOM model has also been evaluated across varying dataset sizes. Our work will serve as a benchmark as well as a reliable technique for decision makers in banking and finance departments for creditworthiness and performance assessments, solely intended for avoiding potential bankruptcy.