- 期刊
學習曲線分析於電腦化動態評量之應用
Application of Learning Curve Analysis to Computerized Dynamic Assessments
摘要
學習曲線主要是用來描述學生學習的次數與答錯率之間的關係，透過學習曲線分析可以幫助教師了解學生的學習狀況，從而更有效的調整教學教材、進行教學，以及提供個別化的指導，因此學習曲線分析在教育領域中是相當重要的。目前，已有多種應用於學習曲線分析的統計模式被提出，要如何從多種模式中選出最貼近學生實際學習情況的模式，以提高分析效率與精準度，為本研究之目的。本研究使用較常見的統計模式（AFM、iAFM、BKT、iBKT），透過模擬方式產生不同的學習曲線，並分析實徵資料，進行模式比較，選擇出最適合的統計模式，並提出使用建議。研究結果顯示，AFM及iAFM的估計精準度比BKT及iBKT高，AFM及iAFM的估計誤差值相近，而iAFM可以估計學生個人的學習率，更能讓教師針對不同學生的學習狀況加以輔導，提高其學習成效。此外，不同的模式表現不會受到起點行為所影響；但在理想的學習曲線中，模式表現會受到學習率的影響。本研究亦透過電腦化動態評量之實徵資料進行驗證與分析，並做為後續研究者或教師在學習資料分析上的一個範例。
並列摘要
Learning curve is applied to describe the relationship of the number of opportunity and the error rate of student responses. Learning curve analysis can assist educators in understanding students' learning situations, thus enabling more effective adjustments to teaching materials, conducting instruction, and providing personalized guidance. The purpose of this research aimed how to select the best model matching the actual learning situation of students' and improving the efficiency and accuracy of the analysis. There were four statistical models, AFM, iAFM, BKT, and iBKT, employed to generate various learning curves through simulation, in addition, our study analyzed actual data to decide the optimal statistical model among them. The research findings indicate that AFM and iAFM exhibit higher estimation accuracy compared to BKT and iBKT. The estimation errors of AFM and iAFM are similar; however, iAFM has the capability to estimate individual student learning rates, enabling teachers to provide tailored guidance based on varying student learning conditions and enhancing overall learning outcomes. Furthermore, the performance of different models is not influenced by students' initial ability; however, in an ideal learning curve, model performance is influenced by students' learning rates. Furthermore, this study validates and analyzes through empirical data from computerized dynamic assessment, serving as an exemplar for subsequent researchers or educators in the analysis of learning data.