利用資訊科技於評估使用者的學習效能,已逐日地重要,若是能隨著學習者目前學習的困難程度進行紀錄,包括各個不同概念領域的了解程度,並且依據此紀錄,提供相關程度的測驗,可以讓學習者在學習上更為有效率,並能夠針對各種領域,進行不同程度的加強。 平行試卷意指同樣的參數設定進行多張試卷的出題,在評估使用者的學習程度時,必須有一致性的標準(目標參數的設定),例如:辨識度、測試問題的困難度、各種概念的涵蓋程度等。如何從一個資料量很大的題庫中,選取一組符合多個目標式的試題來組成一張試卷是極為困難的工作,而同時組成多張平行試卷的情況下更是困難,因為各張試卷都必須符合多種目標參數。 在本論文中,提出了符合多目標式的多張平行試卷產生方法,利用多目標粒子族群演算法MOPSO(Multiple-Objective Particle Swarm Optimization)來建立配題機制,找尋近似最佳解,經過大量的實驗測試之後,發現本方法比其他的方法更為有效,更接近最佳解。
Information technology (IT) has been increasingly important in assessing the learning performance of individual learners. In particular, it is helpful to automatically monitor the bottleneck of learning status such as the degree of understanding for every concept, and provide future test sheets that are focused on the concepts the learners lack. For parallel test sheets generation, the assessment factors across multiple sheets should have the same requirement level. For example, the discrimination degree, difficulty degree, coverage degree of relevant concepts, etc. It is thus very difficult to generate parallel test sheets satisfying multiple criteria. In this thesis, we propose a novel parallel test sheet generation method based on multiobjective particle swam optimization (MOPSO). Experimental results show that our method is superior to a genetic algorithm-based method and the non-dominated solutions found by our method is closer to the true Pareto-optimal front.