Abstract

The present study used Neural Network Backpropagation combined with Nguyen-Widrow to optimize the disadvantages of ANN causes, which is the difficulty in initializing initial weights. The test was conducted on a dataset of values in semesters 1 and 2. The test results show that the best performance in training model of artificial neural networks with Nguyen-widrow is the smallest average MSE error of 0.002 and the highest average accuracy of 96.38%. Training on artificial neural network model training data with Nguyen-widrow has the smallest MSE error, 0.000996 and the highest accuracy is 97.49% on architecture ANN 9-9.1 with training function parameters: traingdx, epoch: 1000, learning rate: 0.1, and error: 0.001. The best performance was also seen in testing the testing of artificial neural network models with Nguyen-widrow with the smallest average error-MSE of 0.026 and the highest average accuracy of 87.85%. Training data testing on artificial neural network models with Nguyen-widrow has the smallest error-MSE which is 0.004436 and the highest accuracy is 94.50% on architecture ANN 9-9.1 with training function parameters: traingdx, epoch: 1000, learning rate: 0.1, and error: 0.001. The artificial neural network model with Nguyen-widrow has an accuracy difference of 8.53% smaller than the artificial neural network model with an accuracy difference of 9.28%. It can be concluded that the Artificial Neural Network with Nguyen-Widrow can overcome the ANN problem in determining initial weights so that it gives an increase in the accuracy of the prediction of students’ competency selection better than the Artificial Neural Network without Nguyen-Widrow.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call