| Artificial Neural network possesses many of advantages: simple structure, easy implementation in hardware, the basic parallel computational architecture, and the model has features of study, memory, self adapting and diversity. Neural network is widely used in more areas ,because it has so many advantages.1. Firstly, the paper introduces the application of artificial neural network, authors deduced the mathematic formula of the single-layer and multi-layer ADALI NE NN module with a hide-layer based on linear transformations. The formula showed that the output of multi-layer ADALINE NN is still liner function accompanying more complex weights and thresholds. It is functional equivalence with single-layer ADALINE NN. So the multi-layer ADALINE NN can be simplified to the single-layer ADALINE NN completely, and the existence of multi-layer ADALINE NN is not essential.2. The paper clarifies the basic principle and the study process of BP neural network and Genetic Algorithms, analyses the existing weakness . Making use of excellent global searching ability of GA and fine learning ability of ANN, the article designs GA-BP Algorithms by using GA to optimize initial weights of neural network. The GA-BP Algorithms, to some extent, can endure the local minima problem widely existed in neural network model training3. Take the score data as an example, we conduct inspection and examination on the score prediction based on GA-BP Algorithms network,apply MATLAB software tosimulation, and compare it with BP Algorithms network,Adaline NN,The result indicates GA-BP Algorithms network has better stability and precision. the average effect of prediction of GA-BP Algorithms is 0.9946, the average effect of prediction of BP Algorithms is 0.9906, the average effect of prediction of Adaline NN is 0.8162. |
