本研究以國立中央大學依「高等教育深耕計畫」所制定之八大課程分類為分析對 象，探討其在學生選課行為中的階段性與關聯性，資料來源為國立中央大學電算中心 校務資料庫總計 2，862，502 筆去識別化之修課紀錄，其中包含民國 111 至 113 學年度的 樣本共 383，322 筆。 研究方法採用描述性統計、時間序列分析、共現分析、馬可夫轉移矩陣、置換檢 定、Louvain 社群偵測與學院別分層分析，藉此辨識課程分類間的縱向階段性變化與橫 向關聯模式。置換檢定結果顯示，相鄰學期間相同課程分類之延續性高於隨機基準； Louvain 社群偵測之正規化互資訊（NMI）結果則顯示，學生實際共修結構與課程分類 之間一致程度有限。 在課程分布分析方面，研究結果顯示「問題導向（PBL）」課程在修課學生數與 課程數上均位居八大課程分類前列，並與多個課程分類呈現較高的共現關係；相較之 下「媒體識讀」課程與多數課程分類之相似度較低，顯示此課程分類在共現結構中呈 現相對獨立的特性。 本研究進一步以學院別進行分層分析，發現不同學院在課程分類的參與情形、年 級階段分布及主要共現組合上均存在差異，顯示全校層級所觀察到的階段性變化與關 聯性模式，可能與各學院之專業屬性及課程結構差異有關。 由於過往的課程分析多著重於單門課程或制度設計上，因此本研究藉由分析學生 實際的修課行為，提供以高教深耕課程分類為主的實證視角，期望研究的成果可作為 校務研究在檢視課程配置與跨類整合時的一項參考依據。;This study takes the eight course categories developed by National Central University under the framework of the Ministry of Education′s "Higher Education Sprout Project" as the object of analysis， examining the stage-wise patterns and associative relationships embedded in students′ course-selection behavior. The data were drawn from the institutional research database maintained by the Computer Center of National Central University， comprising a total of 2，862，502 de-identified course enrollment records， of which 383，322 records correspond to the 2022–2024 academic years (the 111th–113th academic years under the R.O.C. calendar). The research methods employed include descriptive statistics， time-series analysis， cooccurrence analysis， Markov transition matrices， permutation tests， Louvain community detection， and college-level stratified analysis， in order to identify both the horizontal linkages among course categories and the longitudinal patterns of their evolution. The permutation test results indicate that the continuity of identical course categories between adjacent semesters is significantly higher than the random baseline. The normalized mutual information (NMI) results from Louvain community detection， meanwhile， reveal only limited agreement between students′ actual co-enrollment structure and the predefined course category scheme. In terms of course distribution， the findings show that Problem-Based Learning (PBL) courses rank among the top of the eight course categories in both the number of enrolled students and the number of distinct courses， and exhibit relatively high co-occurrence with multiple other categories. By contrast， media literacy courses show relatively low similarity to most other categories， indicating that they remain comparatively independent within the co- iv occurrence structure. This study further conducted a stratified analysis by college， finding that colleges differ in their patterns of course-category participation， grade-level distribution， and dominant cooccurrence combinations. This suggests that the stage-wise changes and associative patterns observed at the university-wide level may be related to differences in disciplinary characteristics and curricular structures across colleges. As prior research on curricula has largely focused on individual courses or institutional design， this study， by analyzing students′ actual course-selection behavior， offers an empirical perspective centered on the course categories of the Higher Education Sprout Project. It is hoped that the findings can serve as a new reference point for institutional research in examining course allocation and cross-category integration.
