Comparisons between Artificial Neural Networks and Fuzzy Logic Models in Forecasting General Examinations Results

MARA Junior Science College (MRSM) Lenggong is one of the educational institutes under Majlis Amanah Rakyat (MARA). Based on the current academic performance and selected criteria of 6A's in the Penilaian Menengah Rendah (PMR, now it is known as PT3), rationally there should be no reason for the failure to achieve excellent results in the Sijil Pelajaran Malaysia (SPM). However, every time the results are announced, the average school achievement grade (GPS) does not meet the performance goals of an average grade of 1.00 for PMR and below 2.00 for SPM, even though it has been in operation for 10 years. Therefore, this research aimed at identifying the influencing factors that affected the students' academic performance. Early prediction is one of the strategies performed in order to improve the students' performance. Neural network and fuzzy logic models are used to realize the accurate prediction based on three factors namely demography, academic and cocurricular activities, including a combination of all three factors. Demography, academic and co-curricular information for the year 2008 to 2010 SPM candidates of MRSM Lenggong are the data sample used. It can be concluded that the prediction outcome using the neural network model shows that the academic factor influences the students' academic performance with the prediction accuracy around 93.65%. Meanwhile, the fuzzy logic model gives an opposite result, where the students' academic performance has also been influenced by the demography factor with an accuracy of 87.00%. Although different techniques yield different results, it is undeniable that the combination of demography and academic factors establishes a solid outcome in identifying the students' present and future academic performances.