Evaluation of quality of teaching based on BP neural network

Evaluation of the quality of teaching, many methods of subjective factors are more computationally intensive, the application up to more trouble for self-learning of neural networks, adaptability, strong associative memory, self-discovery rules and other correspondence between the data advantages, proposed a BP neural network method of evaluation of teaching quality, using the MATLAB neural network evaluation design process. Evaluation of three-layer BP neural network, the hidden layer transfer function tansig, the output layer transfer function purelin. To ensure the network's generalization ability, the paper chooses training samples and more representative selection of data to disrupt the order of pre-normalized data and training function, using Levenberg-Marquardt algorithm, parameters learning rate of 0.01, the network training error of 1e-5 with 1000 training times to choose to step 10. By successfully training the neural network, the paper aims to give four sets of data for a forecast, predicting fitting and error analysis to draw graphics. From the comparison of the data and graphical analysis, the paper has the following conclusion: the successful design of the BP neural network model, based on a reasonable sample of data, the neural network will be through the training samples to learn how to sample the environment intrinsic regularity of the future given the correct input response, successful evaluation of the quality of teachers.