Research on the Optimal Methods of Financial Distress Prediction Based on BP Neural Networks
Xi Zhou, Jiayang Wang, Wei Cheng Xie, Hong Yuan · International Conference on Electric Information and Control Engineering · 2012
The optimization of financial distress prediction system has been studied in this paper, including the input dimensions optimization of neural network using rough-set-based reduction technique, and neural network weights and thresholds optimization using the genetic algorithm, and the results of comparing two optimized models with the conventional BP neural network model. The concrete contents including: rough-set-based reduction technique are used for early-warning index reductions, so that reduce complexities of neural networks and improve network speed and prediction accuracy, genetic algorithm are used as the pre-installation of neural network model to optimized the network input value and the threshold, so as to shorten the network training time and improve the prediction accuracy. Empirical studies have shown that the prediction of the optimized model is more accuracy than traditional BP neural network model.