Modelling Academic Achievement among Selected Public and Private Schools in Ghana: A Bayesian and Artificial Neural Network Approach

Abstract

In the evolving landscape of basic education in Ghana, both public and private basic schools face the challenge of remaining competitive. This study employs Bayesian Loglinear Regression and Artificial Neural Networks (ANN) to identify factors influencing Basic Education Certificate Examination (BECE) outcomes in Ghana's public and private schools.  Data collected from randomly selected schools in two districts in Ghana includes teacher, student, and administrative variables. Bayes factor analysis aids model selection, emphasizing predictive accuracy while balancing complexity. ANN, employing three partitions (6-2-2, 5-3-2, and 7-2-1) for batch training, ranks factors based on normalized importance. The results show superior performance of ANN3, achieving the highest AUC values and efficient generalization to new data. The analysis ranks factors based on normalized importance using ANN, revealing quality supervision, conducive teaching/learning environment, and parents' support as top priorities. These factors significantly impact success rates in both public and private schools. Further investigation highlights the importance of timely provision of teaching and learning materials, consideration of students' interests, and differentiated tasks based on ability levels. Gender and Social Inclusion (GESI) compliance by teachers and the incorporation of critical thinking in lesson delivery also play crucial roles. The study underscores the need for teacher training to enhance critical thinking promotion and the effective use of assessment as a learning strategy. The findings provide a comprehensive understanding of factors influencing academic achievement, offering valuable insights for stakeholders to prioritize and enhance different aspects of the educational system in Ghana.