早期預測與個人化干預：透過線上學習行為與記憶保留程度幫助有風險學生;Early Prediction and Personalized Intervention： Helping At-risk Students via Online Learning Behavior and Memory Retention

NCU Institutional Repository > 資訊電機學院 > 資訊工程研究所 > 博碩士論文 > Item 987654321/89747

請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/89747

題名:	早期預測與個人化干預：透過線上學習行為與記憶保留程度幫助有風險學生;Early Prediction and Personalized Intervention： Helping At-risk Students via Online Learning Behavior and Memory Retention
作者:	陳梓瑤;Chen, Tzu-Yao
貢獻者:	資訊工程學系
關鍵詞:	預測學習分析;個人化干預;遺忘曲線;記憶保留;Predictive learning analytics;Personalized Intervention;Forgetting curve;Memory Retention
日期:	2022-07-12
上傳時間:	2022-10-04 11:58:22 (UTC+8)
出版者:	國立中央大學
摘要:	隨著網路時代的改變，傳統教育逐漸轉化為線上學習，而透過學習數據與人工智慧結合幫助學生學習也是近年逐漸關注的研究議題。本研究以線上學習行為為主軸，研究包括：早期預測與遺忘曲線。透過學生線上自主學習行為來預測學習成效，及早辨識有學習風險的學生，提供個人化干預措施，並探討學生記憶程度與學習成效之間的關聯，與及時干預是否能提升學生學習成效與記憶程度。在早期預測中，本研究利用機器學習分類方法Random Forest搭配ANOVA Univariate Test特徵選取學習行為訓練最佳學習成效預測模型，透過課程設計之程式概念將學習行為特徵轉換為時序性資料，找出最佳干預時間，利用可解釋AI模型-SHAP解釋預測結果，並透過個人化儀表板提供學生早期學習狀況預警與需要加強的學習行為建議。本研究利用實驗探討早期預測與干預對於學習成效是否有幫助，根據結果可知基於學習行為的個人化干預是能有效提升學生學習成效的。另一方面，本研究基於記憶保留，透過學生線上自主測驗行為將學生的學習記憶程度量化，並比較記憶程度與學習成效之間的關聯。研究結果顯示，不同記憶程度分群的學生之間，學習成效會有所差異，對於學習內容記憶程度愈高的學生，其學習成效越好；再者，本研究也發現不同記憶程度分群的學生，經過干預後其學習成效提升會顯著高於缺乏干預的學生。然而，本研究也探討了干預是否會影響學生的記憶程度，根據研究結果，針對學生學習行為的干預對於學生記憶程度的影響較不顯著。 ;With the change of the internet era, traditional education has been gradually transformed into online learning. The combination of learning data and artificial intelligence to help students learn has become an increasing concern in recent years. The purpose of this research mainly focuses on researching early prediction and forgetting curve theory via online learning behavior. Predict students′ learning outcome through online learning behavior, identify at-risk students, provide personalized intervention, explore the relationship between students′ memory and learning outcome, and whether timely intervention can improve students′ learning outcome and memory. In the early prediction, this study uses the machine learning classification method Random Forest combined with ANOVA Univariate Test to train the best prediction model. Learning behaviors are transformed into time-series data through concepts to find the best timing of intervention, and the interpretable AI model - SHAP is used to explain the prediction results. The results provide early warning of students′ learning status and recommendations on strengthening learning behaviors through a personalized dashboard. This research does experiments to explore whether early prediction and intervention are helpful to learning outcomes. According to the results, personalized intervention based on learning behavior can effectively improve students′ learning outcomes. On the other hand, based on memory retention, this study quantifies students′ memory through students′ online testing behavior and compares the relationship between memory level and learning outcome. The results show that there are differences in learning outcomes among students with different memory levels. The higher memory level of learning, the better learning outcome; Moreover, this research found that students′ learning outcomes from different memory levels are significantly higher than that without intervention. However, this research also explored whether the intervention would affect students′ memory. According to the results, the intervention based on students′ learning behavior has a less significant effect on students′ memory.
顯示於類別:	[資訊工程研究所] 博碩士論文

文件中的檔案:

檔案	描述	大小	格式	瀏覽次數
index.html		0Kb	HTML	25	檢視/開啟

在NCUIR中所有的資料項目都受到原著作權保護.

社群 sharing

資料載入中.....