A GA-BP Neural Network Model for Students Examination Score Prediction

Abstract

To improve the teaching effect of "Computer Application Basis" course, this paper selects the students’ behavior data in a Chinese university and uses convolution neural network as the basic model to predict students’ final grades. Considering the disadvantages of traditional B-P neural network model, such as over-reliance on empirical knowledge, unpredictability and difficulty in training, we propose an improved adaptive B-P neural network algorithm based on genetic algorithm. By encoding the B-P neural network, the algorithm takes the calculation error and structural complexity as fitness functions, which can ensure the population diversity of the genetic algorithm and improve the convergence speed, thus effectively avoiding the algorithm from falling into the local optimal solution. To verify the performance of the model in performance prediction, we use MSE, MAE and RMSE to evaluate the performance of the model comprehensively. Experiments show that the convolution neural network optimized by genetic algorithm has achieved good results in calculation accuracy and parameter setting, and is superior to the traditional convolution neural network model.