以分類演算法分析網路學習檔案以增進問題本位式學習的成效

Abstract

近幾年來，問題本位式學習(Problem Based Learning, PBL)以在歐美等先進國家實行有多年之久，並且逐漸由原本所實行的醫學、法學等領域，擴展到各個領域中，並且有著相當不錯的表現，然而PBL的實行過程中，往往需要大量的人力資源，教師需要時常給予學生指示。隨著近代電腦科技蓬勃發展與網際網路不受時空的限制特性的催化下，科技可以更有效率促進學習與減輕老師的負擔，更可較全面的照顧到所有學生。在本研究中，我們嘗試設計PBL的教學活動在開放源始碼之Moodle數位學習平台，進行資訊融入教學活動。而其學習歷程檔案，包含學生在聊天區的聊天內容、討論區的討論內容，wiki的文章相關度等等，而在足可信賴的質量兼具之樣本數提供下，運用現有的分類演算法，進行分析，來鑑別學生的學習成效，讓電腦可以半輔助地模擬教師的角色，幫助引導學生完成問題本位學習模式的學習主題，如此可妥善解決可能的班級教師不足的問題，並且完整真實地紀錄學生的學習歷程，以及在學習的過程中給予適當的評量與鼓勵回饋。在我們以國中兩班進行現場施測後，實驗的結果顯示我們所提出的學習平台的確能有效的輔助學生學習，因此驗證了我們所提出之資料探勘技術的可行性。

Problem based learning (PBL) has been implemented for years in developed countries, such as America and Europe, etc, and extended from original medical science and law to many fields and the achieved performance is plausible. However, the implementation of PBL course often needs a lot of human resources, the teachers often need offering instructions to students. As the modern computer science and the Internet gains wide popularity around the world, elearning is taken by the learners as an important study aid and thereby lightens the burden of the teachers.Recently, a lot of open source elearning platforms can be used for free in the Internet due to the boosting of open source software. In this research, we incorporate the PBL activity into an open software elearning platform, Moodle, and a learning diagnosis and assessment module is added in the platform to alleviate the loading of the teachers. The original data for the experiments will be obtained from the learners’ portfolio recorded during teaching and learning activities which is further fed into a classification algorithm to give an appropriate feedback to the learners. Besides, the learners’ portfolio is also fed into a regression algorithm to obtain the class participation assessment of the learners. The experiment results verify the effectiveness of the proposed algorithms.