研究生: |
劉士華 |
---|---|
論文名稱: |
測驗化知識診斷系統-以類神經網路製作 Testing-based Knowledge Diagnosing System-Neural Network Approach |
指導教授: | 張國恩 |
學位類別: |
碩士 Master |
系所名稱: |
資訊教育研究所 Graduate Institute of Information and Computer Education |
畢業學年度: | 82 |
語文別: | 中文 |
論文頁數: | 69 |
中文關鍵詞: | 測驗化知識診斷 、類神經網路 |
論文種類: | 學術論文 |
相關次數: | 點閱:120 下載:0 |
分享至: |
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報 |
本論文使有類神經網路結構建置迷失概念診斷系統,系統能根據學生的反應來判斷學生的迷失概念。學生答錯問題,可能是有單一迷失概念,亦有可能有多重迷失概念;對於多重迷失概念的區分,系統能決定最恰當的題目來輔助判斷。至於診斷迷失概念所採用的方法,本論文有提出一套新見解,新理論的思考角度跟以往的理論不同,另外亦說明了新理論的實施細節。
為了考驗新理論是否真的能夠提高診斷迷失概念的功效,本論文選擇測試的領域是基本電學。首先分析基本電學領域的迷失概念,並且設計適當的題目。使題目之間的診斷功能相異。然後讓一些高中生作這些題目,根據學生的答案,以面談的方式來確認學生真正的迷失概念。
本論文基本上是設計出一個以類神經網路為架構的通用迷失概念診斷系統,亦即能針對各種不同領域來動態調整神經元數目。系統會根據特定領域的迷失概念集合而自動產生每一層的神經元,然後訓練產生各個神經元間的神經鍵值,這對於通用性而言是很有幫助的。
This thesis uses neural network to build the test - based knowledge diagnosing system. According to the reacton of a student, this system can determine what kind of misconception the student really has. A student might have a single misconception or multiple misconceptions if he picks the wrong answer. And this system can determine the right question to distinguish and diagnose the exact misconception from the different multiple misconceptions. This thesis also suggests a new methodology to diagnose misconceptions. and a new different thinking angle with the detailed executing method.
For examining whether the new methodology could make the diagnosis more efficient, this thesis chooses basic electricity as the tesing domain. To analyze the misconception of the specific domain and then design the appropriate questions to fit the different situation. We had some senior high school students answer the questions that I designed and then talked to them by interviewing to realize their real misconceptions.
Basically, the thesis design a general system to diagnose misconnceptions. And it use the neural network, which can dcynamically adjust the number of neuron in different domains. The system automatically produces neurons on every level according to the set of misconceptions in specific domain and trained to generate the weight. This is helpful to general application.