應用資料探勘技術分析學生選課特性與學業表現

The Application of Data Mining Technology in Analyzing the Characteristics of Student Course-Choosing and Academic Achievement

顏博文(Po-Wen Yen)

指導教授 ： 李維平

中原大學/商學院/資訊管理研究所/碩士(2003年)

https://doi.org/10.6840/cycu200300219

摘要

目前台灣大學教育的發展，逐漸朝向學程制與大一不分系的趨勢邁進，各大
學累積了大量且完整的學生學籍資料與選課紀錄資料庫，其中可能隱藏著許多未
被發現的知識，對於學校課程的設立與規劃有所幫助。本研究使用資料探勘的技
術於學生選課紀錄找出潛藏其中的知識，運用於輔助系主任規劃課程與學生學
習，此外也期望能探勘出影響學生平均學業表現的特質，提供系上招生時的參考。
我們分析中原大學資訊管理學系學生資料，期望發現有用的潛在規則與知
識，提供行政單位做為開課與招生決策的參考。因此，本研究希望達到以下目的：
1、探勘性別、不同入學管道、居住區域等學生學籍資料對學業表現是否有
2、探勘學生選課行為模式間的關聯，並探討其中的特徵。
3、將學生分群，了解不同族群間的特色。
資料探勘此新觀念在近幾年內興起，目前國內的相關研究中以金融、銀行、
保險等企業領域的應用較多，對於教育界算是一個新的應用。透過此次利用集群
分析、決策樹、序列型樣等不同的資料探勘方法，針對學生特質、平均學業表現、
和選修課程序列等方向進行探勘，可以歸納得到下列結論：
一、影響學生學業表現最重要的預測屬性為分別為『性別』，其餘依次為入
二、學生來源多偏重於台北和桃園縣市，桃園縣市附近的苗栗、新竹地區學
生表現普遍優秀，學校可藉地緣關係多加強招生。
三、推薦甄試多集中於北部縣市，系上可以考慮朝向中南部地區學生學業表
四、資管系的學生普遍喜好選修多媒體課程、其他課程，且不僅只修一次課
程，系上在規劃此類課程，可考慮多開相關課程。

關鍵字

決策樹 ； 學業表現 ； 資料探勘 ； 序列型樣

並列摘要

For the time being, the development of the university education in Taiwan is
moving toward the trend of "Program" and "core curriculum". Each university and
college has accumulated huge and complete student databases, which may hide lots of
undiscovered knowledge. Besides, this knowledge can be useful for building and
planning the courses. The study, which is explored the hidden knowledge by data
mining technology, can help the staff of the department to design the courses, and is
expected to find the characteristics related to the students' academic achievement, for
the purpose of being a reference whe n enrolled.
By analyzing the students' profile in the Department of Management Information
Systems in CYCU, we try to discover potential rules and knowledge, and provide the
administration for delivering courses and recruiting strategies. Thus, the study aims at
1. Mining if the academic achievement is predictable by researching students'
profile, such as sex, different methods of entrance and resident regions.
2. Mining students' course-chosen patterns and discussing the characteristics.
3. Clustering students and analyzing the differences among those clusters.
Data mining is a new concept rose in recent years. Most of the studies bound up
with data mining are focused on Finance, Banking and Insurance, as for Education,
data mining is a new application. By using different data mining methods, such as
clustering, decision tree, and sequence pattern, and analyzing students' characteristics,
average academic achievement and course-taken sequence, we can list the conclusion
1. The most significant factor which can predict academic achievement is "sex",
and the rest in turn are entrance methods, resident areas and the schools from
2. Most of the students come from Taipei and Tao-Yuan. Besides, students who
come from Miao-Li and Hsin-Chu have better academic achievement. Therefore,
we suggest that the administration can do more recruiting activities in these areas.
3. Most of the students who admitted via screening and recommendation entrance
system come from the North Taiwan. Thus, we suggest that the administration can
recruit students who have better academic performance from the Central and the
4. Many MIS majored students prefer to add more than one multimedia related
course and other courses which are not related to Information Systems. Hence, we
suggest that the administration can plan more courses mentioned above.

並列關鍵字

Decision Tree ； Acdamic Achievement ； Data Mining ； Sequential Pattern

參考文獻

[4] 陳正道，『影響學生學業表現相關因素之探討—以亞東技術學院工業管理

生』，南台科技大學資訊管理研究所會議論文，2001

[1] 朱克剛、夏延德，『由選課紀錄推斷課程分類在心目中的重要性排序』， 中

[2] 梁俊明，『ART 在教育資料探勘的應用』， 國立臺灣師範大學資訊教育研

establishing a Tiered Knowledge Management Model",2001