Application of Simulated Annealing Algorithm in the Construction of Online Examination System for Tax Law Courses

The increasing reliance on online examination systems in tax law education necessitates the development of intelligent frameworks that ensure fairness, syllabus coverage, and difficulty balance. Traditional exam construction methods, such as manual question selection and heuristic approaches, often result in inconsistencies, bias, and suboptimal difficulty distribution. This study proposes a simulated annealing (SA)-based optimization model for dynamically generating balanced online examinations in tax law courses. The SA algorithm efficiently selects exam questions while minimizing Exam Difficulty Variance (EDV), maximizing Syllabus Coverage Ratio (SCR), and reducing computational overhead. Experimental results demonstrate that SA achieves superior performance compared to genetic algorithms (GA), greedy approaches, ant colony optimization (ACO), and particle swarm optimization (PSO). Specifically, SA attained the lowest EDV (0.038), the highest SCR (94%), and the fastest execution time (ET) (1.5s), outperforming GA (5.2s) and ACO (3.8s). These findings highlight the effectiveness of SA in generating equitable and adaptive exams, providing a scalable and automated solution for tax law education. Future enhancements include integrating machine learning techniques to refine SA’s parameter tuning and further improving exam personalization.