- 學位論文
應用群集分析法於課堂計畫分組之研究
Applications of Cluster Analysis on Class Project Grouping
摘要
老師的教學策略中，課堂分組是常被使用來幫助學生提升學習能力的方法之一，通常為了分出同質性的組別，多半是依據過往教學經驗或是直觀感覺，有時無法有效且快速地幫助學生分組，也缺乏分組操作上的建議依據，所以本研究利用群集分析法(Cluster Analysis)裡的階層式群集分析(Hierarchical clustering)與非階層式群集分析(Non-hierarchical clustering)來求解分組結果，透過學習風格量表(Index of Learning Style)數據作為分組的參考數值，找出相同學習風格的學生為一組，並藉由常態分配模擬生成大量在不同總人數、組數與參數等條件的分組錯誤率，作為往後使用的參考依據，最後應用於實例上，比對群集法的結果與人工的結果來驗證實用性。透過模擬資料與實例結果顯示，完全法(complete method)、華德法(ward method)與k中心點法(k-medoids method)此三種群集法運用在學習風格量表數據上，有良好的分群結果，另外也發現人數提高，分組錯誤率也跟著提高，最後在實例應用上，群集法分組結果與人工分組結果是明顯差異。
並列摘要
Class grouping in instructor’s teaching strategies are usually used for improving students’ learning abilities. To group homogeneous grouping, the way is usually relied on past teaching experience and intuitive feeling. So sometime the way don’t help student grouping well and immediately. And also lacking suggestions give to users. Therefore, in the research, we use Cluster Analysis (CA)’s hierarchical clustering and non-hierarchical clustering to solve the grouping results. Using Index of Learning Style (ILS) data become grouping reference data and find homogeneous groups. And using CA’s suggestions in future are relied on a large number of grouping incorrect rate data which generate from normal distribution in different total people numbers、grouping numbers and parameters. Finally, by using in real case, we compare CA’s grouping results and human grouping results to verify practicality. The simulation data and real case‘s results show complete、ward and k-medoids methods have great grouping results in using ILS data. And, when total people numbers increase, grouping incorrect rate also increase. Finally, using on real case, CA’s grouping results is clearly different with human grouping results.