隨著電子化教學媒體的快速發展，遠距教學環境逐漸從純電子書式的瀏覽進展為開始關心每一個學習者的個別需求。適性化教學的個人化學習特色，正是一個可以滿足現今遠距教學環境的教學方法。要達成適性化的教學目標，需要偵測學習者的學習狀態並產生相對應的個人化教材連結。對於程式語言的教學而言，熟練度是一項典型的指標，可以用來代表一個程式語言學習者的學習狀態。然而，在傳統的觀念中，熟練度過於抽象，導致我們只能知道熟練的人所寫的程式，比較能兼具效率、效能與品質；卻無法描述熟練度的值並進行比較。 因此本論文從程式語言學習與教育學的角度出發，探索一個程式語言專家所具備的行為特質，並依此提出一個熟練度的量化模型。根據量化參數，我們可以得到一個熟練度指標，用來標示每一位程式學習者的學習狀態。我們使用了一個線上學習系統(NMPTE)來收集學習者的學習行為。藉由線上學習資料與傳統考試結果的相關性分析，可以說明我們所提出來的熟練度模型足以代表程式語言的學習成效。因而可以輔助線上學習系統找到個人化的學習路徑，以達成適性化教學的理想。 With the rapid development of e-learning, education is shifting the paradigm from the pure electronic browsing to taking care of each learner’s learning state. Adaptation is an ideal learning framework to accommodate such a requirement. The strategies of detecting learning state and generating relative adaptation links affect the success of adaptive navigation. Proficiency is a classic indicator in testing learning states of programming. Similar to data mining, proficiency mining is added in the learning process to explore each learner’s proficiency from observable behavior. This dissertation presents a web-based learning system (NMPTE) embedded with a proficiency mining strategy that is used for supporting adaptive navigation. The strategy involves using quantified parameters to characterize the process of coding rehearsal. The criteria of selecting parameters are based on the psychology researches and the basic properties of learning to program. Web technologies with the standard of XML are adopted to describe adaptive curricula and proficiency parameters. Our experience demonstrates that proficiency mining shows sufficient evidence to represent programming performance and becomes a tool to detect users’ features in adaptation system.
