數據特徵選取結合適應性增強演算法於學生數學與科學成就之預測

Adaboost with ReliefF in predicting students’ mathematics and science achievement

洪鈺梅(Yu-Mei Hung)

指導教授 ： 白炳豐

國立暨南國際大學/管理學院/資訊管理學系/碩士(2011年)

摘要

適應性增強演算法為一種新興演算法，此演算法用於降低學習性演算法之錯誤率。然而，對於國際數學與科學教育成就趨勢調查之議題應用於適應性增強演算法尚未被廣泛地討論。此外為了提升效率並減少時間計算複雜度，本研究採用數據特徵選取及後處理方法。被採用前者方法用於定義重要的訊息屬性，後者方法則是剔除部必要之離群值。運用前處理及後處理程序可減少不當的決策，因此，本研究提出得以結合數據特徵選取及適應性增強演算法之模型以提供使用者預測學生之數學與科學成就之重要屬性。此外本研究更延伸探討學生與組織文化之關係性，此論點可有效提供給教育學家或家長做有效的決策。

關鍵字

數學與科學研究 ； 國際數學與科學教育成就趨勢調查 ； 數據特徵選取 ； 特徵選取 ； 適應性增強演算法 ； 組織文化

並列摘要

Adaboost is an emerging algorithm used to reduce the error rate of learning algorithm. However, there are not wide discussion about the application of Adaboost for predicting students’ mathematics and science achievement from the Trends in International Mathematics and Science Study (TIMSS). Furthermore, to improve the efficiency and reduce the computational complexity, the study employs the ReliefF algorithm and the post-processing. The former was utilized to determine the important informative features, and the latter is conducted to omit the outlier. Applying the pre-processing and post-processing procedures would eliminate improper decisions. Hence, the research proposed ReliefFAda model which incorporate ReliefF with Adaboost mechanism to predict important attributes for users to analyze the students’ mathematics and science achievement. Moreover, the study further discussed the organization culture in student’s relations. It is very useful for educationist or parents to make a reference and make a suitable decision.

並列關鍵字

Mathematics and Science Study ； TIMSS ； ReliefF ； Feature selection ； Adaboost ； Organization culture

參考文獻

[1] Andreas Ammermu¨ller, Hans Heijke, Ludger Wo¨Xmann. Schooling quality in Eastern Europe: Educational production during transition.Economics of Education Review 24, 579–599, 2005.

[2] Arjun Chandra, Xin Yao. Evolving hybrid ensembles of learning machines for better generalization. Neurocomputing 69, 686–700, 2006.

[3] Aziz Guergachi, Goran Boskovic. System models or learning machines. Applied Mathematics and Computation 204, 553–567, 2008

[4] Betiil Yayan , Giray Berberoglu. A re-analysis of the TIMSS 1999 mathematics assessment data of the Turkish students. Studies in Educational Evaluation 30 87-104, 2004.

[5] Blake, C. L., Merz, C. J. UCI repository of machine learning databases. http://www.ics.uci.edu/mlearn/MLRepository, 1998