Comparative Analysis Various Membership Functions Based on Neural Networks for Prediction of Students Graduation

It is very important to be able to predict student success and understand the factors that affect academic performance in order to improve graduation rates and the quality of education. This study used ANFIS to predict student graduation using different membership functions, including triangular, trapezoidal, Gaussian, and generalized bell membership functions. The study also analyzed how the results differed for each membership function. The Gaussian membership function performed the best on both datasets, with an average MSE of 0.032893, an average RMSE of 0.17477943, an average correlation coefficient of 0.93516, and an average determination coefficient of 0.876332175. While the other models still performed well overall, the ANFIS model with the trapezoidal membership function performed the worst, with the only model having a determination coefficient of less than 0.6. Therefore, the ANFIS model with the Gaussian membership function is recommended as a reliable model for making predictions.