根據SHAP解釋模型提供基於Coding pattern干預以提升學習成效;Enhancing Learning Effectiveness through Coding Pattern-based Interventions Provided by the SHAP Explanation Model

NCU Institutional Repository > 資訊電機學院 > 資訊工程研究所 > 博碩士論文 > Item 987654321/93015

jsp.display-item.identifier=請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/93015

题名:	根據SHAP解釋模型提供基於Coding pattern干預以提升學習成效;Enhancing Learning Effectiveness through Coding Pattern-based Interventions Provided by the SHAP Explanation Model
作者:	林文涵;Lin, Wen-Hen
贡献者:	資訊工程學系
关键词:	Coding pattern;程式初學者;干預輔導活動;SHAP;Coding pattern;Programming novice;Intervention;SHAP
日期:	2023-07-05
上传时间:	2024-09-19 16:38:31 (UTC+8)
出版者:	國立中央大學
摘要:	資訊科技日益普及，程式能力逐漸成為教育當中不可或缺的一部分，大學也開始發展多種資訊科學相關的課程，並在課程中使用各種系統作為學習環境，像是電子書系統、程式編譯環境、開放交流論壇等，讓學生可以不受到時間和地點的限制來學習程式。透過早期的研究可知，程式初學者容易有程式語法以及邏輯方面的學習困擾，而且可能無法自行從錯誤訊息中發現並解決，最終導致程式設計任務失敗，因此如何識別出學生的學習情況以及了解學生的行為面向並給予干預回饋成為重要的議題。本研究從程式初學者的線上程式設計行為中識別出Coding patterns推斷學生的行為面向，以及計算早期研究使用過的程式設計統計數據作為特徵，進行學習成效預測，預測方法是使用機器學習方類方法，透過常見的預測模型評估因子找出最佳預測學習成效的模型。接著使用SHAP計算預測模型中Coding patterns的影響不同預測結果之程度及方向，亦即對於預測結果的貢獻度，最後設計出學生個人化干預內容，並呈現在學生個人干預儀表板作為基於Coding pattern的干預輔導來提升學生之學習成效，最後探討此個人化干預輔導的有效性。;Information technology is increasingly prevalent, and programming skills have gradually become an essential part of education. Universities have also started developing various courses related to computer science and using various systems as learning environments, such as e-book systems, programming compilation environments, and open communication forums, allowing students to learn programming without being restricted by time and location. Early research has shown that novice programmers often encounter difficulties in learning programming syntax and logic and may struggle to identify and resolve errors on their own, ultimately leading to failed programming tasks. Therefore, it is important to identify students′ learning situations and understand their behavioral aspects in order to provide intervention feedback. In this study, coding patterns were identified from the online programming behavior of novice programmers to infer students′ behavioral aspects. Early research used programming statistics as features to predict learning outcomes. Machine learning methods were employed to predict learning effectiveness, and common prediction models were evaluated to find the best model for predicting learning outcomes. The SHAP (Shapley Additive Explanations) method was then used to calculate the influence of coding patterns in the prediction model on different prediction results, that is, their contribution to the prediction results. Finally, personalized intervention content was designed based on coding patterns and presented in a student′s personalized intervention dashboard to improve their learning outcomes. The effectiveness of this personalized intervention guidance was subsequently investigated.
显示于类别:	[資訊工程研究所] 博碩士論文

文件中的档案:

档案	描述	大小	格式	浏览次数
index.html		0Kb	HTML	21	检视/开启

在NCUIR中所有的数据项都受到原著作权保护.

社群 sharing

数据加载中.....