臺灣博碩士論文加值系統

近年來，大數據的發展趨勢使得校務研究逐漸成為眾多學校關注的議題。為了應對這一趨勢並提高教學品質，本校成立了校務研究單位，整合了包括學生學業成績、課程選擇、社團參與等多個維度的數據，構建了一個既豐富又複雜的資料倉儲。然而，校務研究員在針對不同主題時，如何從資料倉儲中組織出適合的資料市集仍是一項挑戰。僅依靠經驗或相關性可能會產生看似有意義但實質無意義的訊息。本研究認為，資料市集的索引維度應該反映出具有因果可能的分析角度，避免不嚴謹的資料市集設計導致分析結果的不準確和難以解釋，進一步影響決策支援的效果。
本研究採用基於相關性特徵選擇（Correlation-based Feature Selection, CFS）來計算和評估特徵集合的價值，並搭配使用向前選擇（Forward Selection, FS）作為具體的特徵選擇方法，篩選出符合特定主題的特徵集合。隨後，透過因果勝算比探勘技術，針對特定主題進行深入分析，同時評估在給定的數據範圍內，該主題是否具有深入探索的可適性。本研究以校務資料倉儲作為資料來源，分別對 「在學適應良好」、「在學不適應」、「多元學習」三個不同主題進行探討，推薦在特定主題中能夠突顯因果相關的資料市集所需的索引維度。藉此協助校務研究人員在教育方針上能更精準且有力，達到決策支援。

In recent years, the trend of big data has gradually made institutional research a topic of concern for many schools. To cope with this trend and improve teaching quality, our school has established an institutional research unit, integrating data from multiple dimensions including student academic performance, course selection, and club participation, forming a rich and complex data warehouse. However, for institutional researchers addressing different topics, how to organize suitable data marts from the data warehouse remains a challenge. Relying solely on experience or relevance may generate seemingly meaningful but essentially meaningless information. This study argues that the index dimensions of the data marts should reflect potentially causal analysis perspectives. Avoiding imprecise data mart design is crucial as it can lead to inaccurate analysis results and difficulties in interpretation, further affecting the effectiveness of decision support.
This study employs the Correlation-based Feature Selection (CFS) method to calculate and evaluate the value of feature sets. In combination with Forward Selection (FS), it is used to filter out feature sets that align with specific themes. Subsequently, using causal odds ratio mining techniques, it conducts in-depth analysis on specific topics while assessing whether the topic is suitable for in-depth exploration within a given data range. This study uses the institutional research data warehouse as the data source and discusses three different topics: "good adaptability in school," "poor adaptability in school," and "diversified learning." It recommends the index dimensions required for data marts that can highlight causal relevance in specific topics. This assists institutional researchers in being more precise and effective in formulating educational policies, thereby achieving decision support.

摘要 i
ABSTRACT ii
誌謝 iv
目錄 v
圖目錄 vii
表目錄 viii
一、緒論 1
1-1. 研究背景與動機 1
1-2. 研究目的 2
二、文獻探討 3
2-1. 資料倉儲（Data Warehouse ） 3
2-2. 資料市集(Data Mart) 3
2-3. 基於相關性的特徵選擇(Correlation-based Feature Selection ,CFS) 4
2-3-1. 向前選擇法(Forward Selection ,FS) 5
2-4. 關聯規則探勘 (Association Rule Mining) 6
2-4-1. 支持度(Support) 6
2-4-2. 信賴度(Confidence) 7
2-5. 因果勝算比探勘(Causal Odds Ratio Mining) 7
2-6. 隊列研究(Cohort Study) 8
2-7. 勝算比(Odds Ratio) 9
三、研究方法 11
3-1. 系統架構 11
3-2. 資料來源與前處理 11
3-3. 基於相關性的特徵選擇流程 13
3-4. 系統流程 14
3-5. 因果勝算比規則候選與驗證 16
3-6. 分析與推薦資料市集所需的維度 17
四、實驗 19
4-1. 實驗環境與規格 19
4-2. 實驗資料集與設計 19
4-3. 實驗分析 20
4-3-1. 主題-在學適應良好 20
4-3-1-1. 因果勝算比規則分析與討論-在學適應良好 22
4-3-2. 主題-在學不適應 25
4-3-2-1. 因果勝算比規則分析與討論-在學不適應 28
4-3-3. 主題-多元學習 31
4-3-3-1. 因果勝算比規則分析與討論-多元學習 32
4-4. 實驗結果統計 33
五、結論與討論 35
六、參考文獻 37

[1] J. L. Saupe, "The functions of institutional research," 1990.
[2] S. T. March and A. R. Hevner, "Integrated decision support systems: A data warehousing perspective," Decision support systems, vol. 43, no. 3, pp. 1031-1043, 2007.
[3] C. Ghezzi, "Designing data marts for data warehouses," ACM Transactions on Software Engineering and Methodology (TOSEM), vol. 10, no. 4, pp. 452-483, 2001.
[4] J. Li et al., "From observational studies to causal rule mining," ACM Transactions on Intelligent Systems and Technology (TIST), vol. 7, no. 2, pp. 1-27, 2015.
[5] T. Dasu and T. Johnson, Exploratory data mining and data cleaning. John Wiley & Sons, 2003.
[6] M. A. Hall, "Correlation-based feature selection of discrete and numeric class machine learning," 2000.
[7] J. Pearl, "Causal inference in statistics: An overview," 2009.
[8] W. H. Inmon, "What is a data warehouse," Prism Tech Topic, vol. 1, no. 1, pp. 1-5, 1995.
[9] A. Jović, K. Brkić, and N. Bogunović, "A review of feature selection methods with applications," in 2015 38th international convention on information and communication technology, electronics and microelectronics (MIPRO), 2015: Ieee, pp. 1200-1205.
[10] J. Hipp, U. Güntzer, and G. Nakhaeizadeh, "Algorithms for association rule mining—a general survey and comparison," ACM sigkdd explorations newsletter, vol. 2, no. 1, pp. 58-64, 2000.
[11] K. Z. Mao, "Orthogonal forward selection and backward elimination algorithms for feature subset selection," IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), vol. 34, no. 1, pp. 629-634, 2004.
[12] P. Sedgwick, "Retrospective cohort studies: advantages and disadvantages," Bmj, vol. 348, 2014.
[13] M. Scriven, "A summative evaluation of RCT methodology: An alternative approach to causal research," Journal of multidisciplinary evaluation, vol. 5, no. 9, pp. 11-24, 2008.
[14] M. Szumilas, "Explaining odds ratios," Journal of the Canadian academy of child and adolescent psychiatry, vol. 19, no. 3, p. 227, 2010.
[15] B. Singh, N. Kushwaha, and O. P. Vyas, "A feature subset selection technique for high dimensional data using symmetric uncertainty," Journal of Data Analysis and Information Processing, vol. 2, no. 4, pp. 95-105, 2014.
[16] A. P. U. Siahaan, A. Ikhwan, and S. Aryza, "A novelty of data mining for promoting education based on FP-growth algorithm," 2018.