Construction and Application of AI-driven Test Question Bank: Integrating Genetic Algorithm to Achieve Differentiated Output

Question bank management faces persistent challenges, including insufficient question volume, labor-intensive manual generation, difficulty balancing, time constraints, and limited knowledge coverage. To address these issues, this study proposes an integrated framework combining artificial intelligence (AI) and genetic algorithm (GA). First, a domain-specific pre-trained large language model (Qwen-Turbo) generates 1,000 questions with automatic feature annotations (ID, difficulty, time, and knowledge points). Subsequently, the GA iteratively optimizes the exam-building process by selecting question combinations that maximize fitness based on constraints such as total difficulty, time allocation, and coverage. Computational results on the PyCharm platform demonstrate that GA achieves stable convergence within 1 second per 100 iterations, with fitness values exceeding 92% of the ideal threshold. T-tests validate the model's effectiveness from both teachers' and students' perspectives. Data indicate that the model significantly reduces teachers' workload (p<0.05) compared to traditional methods, while differentiated exam papers better meet students' learning needs (p<0.05). This study proves the efficacy of AI-driven question banks and GA-based test paper generation in enhancing teaching quality, offering a scalable solution for developing differentiated test papers.