Artificial Neural Network Approach for Predicting Student Achievement in Scratch Training

Abstract

In the contemporary era, the acquisition of vital competencies for the twenty-first century, such as critical thinking, analytical reasoning, and problem-solving, has assumed paramount importance for individual sustenance. A substantial corpus of scholarly research has surfaced, with a primary emphasis on utilizing machine learning and statistical methodologies to forecast the academic achievement of pupils in higher education. Nevertheless, the importance of predicting the academic achievement of elementary and middle school pupils, specifically in the field of computer programming, is growing significantly. The primary objective of this study was to utilize an Artificial Neural Network (ANN) for the purpose of predicting student outcomes in a trial. These predictions made by the ANN were subsequently compared against expert evaluations. Furthermore, the primary objective of this study was to examine the hypothesis that the level of students’ interest in coding and algorithms can be enhanced by the creation of games using a visual programming platform designed for beginners. In order to achieve this objective, the Artificial Neural Network (ANN) was employed to predict the performance of middle school children in a programming exercise. The activity was done on the Scratch platform, and involved the participation of three separate cohorts of middle school students from Turkey. The best results were obtained using the Bayesian regularization algorithm: Training-R = 9.5797−1; Test-R = 7.5072−1, and All-R = 9.3752e−1. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.