Modelling Mathematical Performance Using Two Staged Hybrid Structural Equation Modelling and Artificial Neural Network Among Secondary School Students

Abstract

Predicting students’ performance is absolutely essential in both educational and machine learning contexts. The objective of the study is to determine the variables which are students’ interest, teachers personality, peer influence and perceived technology that can give the significant effect on students’ performance on mathematics subject through the structural equation modelling (SEM) and artificial neural network (ANN) combination. A sample comprised 302 respondents selected from a certain school in Malaysia using convenience sampling. The instrument used is online questionnaire that is validated by using exploratory factor analysis (EFA) for pilot study and confirmatory composite analysis (CCA) for field study. The model was validated using a multi-analytic approach using SEM and the results from SEM were used as inputs to the ANN model to develop a predictive model. The results of model A indicated that geometry domain has the greatest influence on students interest that can affect students score in mathematics. It is followed by discrete mathematics, statistics and probability, number and operations and algebra domain. The results of model B indicated peer influence was the most significant factor that can influence students score in mathematics, followed by students interest, perceived technology and teachers personality. The findings of this study can be critical and valuable for the school in predicting students’ performance in mathematics subject. Hence, the results of this study can be used to assist teachers to formulate strategies as early as possible and to cover the stimuli affecting students’ performance with appropriate measures as preparation for the students to sit for the national examination, Malaysian Certificate of Education (SPM).