What Is Neural Network Analysis?
The neural network analysis method is a processing method with a high degree of parallel computing ability, self-learning ability, and fault tolerance developed from the research results of neuropsychology and cognitive science.
Neural network analysis
- What is neural network analysis
- Neural network technology in pattern recognition and classification, recognition filtering, automatic control,
- Application of neural network analysis methods
- When neural network analysis is used in the study of corporate financial conditions, it uses its mapping capabilities on the one hand, and its generalization capabilities on the other. And interpolate and extrapolate the data in the new situation to infer its attributes.
- Neural network analysis to predict financial crisis Although the theory of neural networks can be traced back to the 1940s, the application of credit risk analysis began in the 1990s. Neural network is a parallel distributed pattern processing system developed from the research results of neuropsychology and cognitive science and applied mathematical methods. It has a high degree of parallel computing ability, self-learning ability, and fault tolerance ability. The structure of the neural network is composed of an input layer, several intermediate hidden layers, and an output layer. Foreign researchers such as Altman, Marco, and Varetto (1995) have applied neural network analysis to the forecast of Italian companies' financial crisis. Coats, Pant (1993) used neural network analysis to predict the financial crisis of American companies and banks, respectively, and achieved certain results. However, the biggest disadvantage of neural networks is that their work is more random. Because to get a better neural network structure, it needs to be artificially debugged, which consumes manpower and time, so the application is limited. Altman (1995) concluded in a comparative study of neural network methods and discriminant analysis methods: the application of neural network analysis methods in credit risk identification and prediction does not substantially outperform linear discriminant models. But neural networks, as a new information processing science, still attract researchers in many fields.