招生為高教機構培育人才之關鍵，如何遴選有潛力之優秀學生成為重要研究議題。本研究運用資料探勘決策樹之J48演算法分析台灣某研究所2008-2012入學共247位學生歷史資料(個人資料、入學考成績、甄試成績、修課成績、畢業成績)，並依分析結果建立學業成績預測模型與招生決策支援系統。研究結果顯示甄試生入學後學業表現較考試生優秀，而男性在入學考試上有較佳表現，女性則在甄試上表現較佳; 私立大學畢業生在入學後的學業表現優秀的機會較國立大學畢業生高。本研究共建立十組學業成績預測模型，其準確率介於23%-67%，預測效果最佳的自變項組合為學生基本資料加上其修課成績。本研究依據預測模型建立招生決策支援系統，使用者可使用此系統輸入資料學生資料，系統會預測輸入學生之未來學業表現，系統功能包括選擇預測功能、輸入資料、學生表現預測、輸出結果、維護系統等五步驟。整體而言，學生的個人資料、考試與甄試成績對其畢業成績之預測正確率不高，顯示目前高教招生制度仍有改善空間。綜言之，現行高教招生制度下，招生過程中成績優異的學生入學後不必然會有同等優秀表現。本研究結果可協助高教機構改善其招生策略。;Student enrollment is an important task for educational institutions in higher education. This study used J48 algorithm to analyze 247 students who enrolled at a university of northern Taiwan between 2008 and 2012. Based on the results, this study built student academic performance prediction models and an enrollment decision support system. The results showed that the students being recruited via the recommendation route performed better than those being recruited via the entrance examination. Moreover, the results also suggested that male had a better performance in entrance examination, while female performed better in recommendation route. This study established ten decision tree models in predicting student academic performance, with success rate between 23% and 67%. The most effective combination of predictor variables is the student demographics and the student grades. This study then used the prediction models to build an enrollment decision support system. One can use the system to predict a student’s academic performance. In conclusion, the personal information and enrollment data of students may not predict academic performance of students correctly. The result suggested that the current enrollment policies and procedures still has room to improve. Overall, the students, who performed excellent in recommendation route or entrance examination, may not perform as well as they previously did after admission. The results can help educational institutions improve their recruiting strategies.