Application of Back Propagation Neural Network and ANFIS in forecasting university program

Abstract

In recent years, intelligent techniques like Artificial Neural Network (ANN) and Fuzzy Logic (FL) have been adopted and tested to solve the real-world problems. The combination of those two powerful approaches has resulted in the emergence of Adaptive Neuro-Fuzzy Inference System (ANFIS). ANFIS itself has achieved a rapid growth in its numbers of applications especially in control system, decision making, classification, forecasting, and modeling. Thus, in this paper, a study has been conducted to analyze and compare ANFIS and BPNN approaches in order to classify the SPM results for diploma programs in Universiti Teknologi MARA (UiTM), Malaysia. Each model has three input data from the public examination, Sijil Pelajaran Malaysia (SPM) results. They are clustered into three parameters, that are, sciences, mathematics and others and one output which is the academic marks for respective UiTM diploma programs. The Root Mean Square Error (RMSE) of the data set is computed in order to observe the accuracy of both models using the same data set. The testing RMSE for ANFIS and BPNN models are 5.6861 and 38.2040 respectively. Consequently, the findings indicate that ANFIS model can successfully forecast the UiTM diploma programs according to their SPM results better than BPNN.