A Machine Learning-Based Automated Scoring Approach for Operator Code Assignments

Automated intelligent code assessment systems for operator training enable operators to receive timely feedback and improve their skills. Recently, the most widely used automated intelligent assessment system for operator training is based on the dynamic test-based method. However, the approach fails to evaluate operators’ programming logic or provide insight into their programming abilities. To address this problem, we propose a machine learning-based automatic code assignment scoring method, which extracts grammatical, semantic, and Control Flow Graph (CFG) information from operators’ code and then uses a classifier to score it automatically. Moreover, we develop an automated tagging approach that utilizes the distance between operator performance and expectations to address the bias of human grading. This makes the system’s scoring more objective and reasonable, providing more effective positive feedback for operators. We construct a dataset containing 48,530 C/C[Formula: see text] source codes, and the experimental results show that our method can achieve promising results with Bias, MAE and Pearson correlation coefficients (r) of 0.04, 0.20 and 0.95, respectively. Furthermore, when comparing our method with the state-of-the-art baselines, we find that our method can outperform them significantly.