教育是弱勢家庭得以擺脫貧困的重要方式之一，然而隨著貧富差距變大，而國家資源亦有限的情形下，弱勢青少年的學業表現則成為重要的議題。本研究運用資料探勘方式探討影響弱勢青少年學業表現之重要影響因素，期提供學校、父母與相關單位參考，以預先早期介入並採取因應措施，避免青少年因學習表現惡化而衍生相關問題。
本研究使用台灣家扶基金會「台灣貧窮兒少資料庫」2015年弱勢兒少生活趨勢調查資料中之國中生問卷(1,289份)進行實驗，實驗採用資料探勘技術方式預測影響弱勢青少年學業表現的重要因素，並運用單純貝氏、JRip及決策樹分類演算法建立預測模型並進行預測模型效能評估。
本研究結果發現針對受調查之弱勢青少年中，影響學業表現重要因素為自我期望受教育程度、能否聽懂學校老師教授的課程、自覺是否用功，預測模型表現則以JRip最佳。
Education is an important way for disadvantaged families to pull themselves out of poverty. However, as the gap between the rich and poor increases while government resources remain limited, the academic performance of disadvantaged teenagers becomes an important concern. In this study, data mining techniques are adopted to examine the key factors that influence the academic performance of disadvantaged teenagers in the hope of providing references for schools, parents and concerned units, so that countermeasures can be taken early to prevent problems derived from deterioration of the academic performance of such teenagers.
The 2015 questionnaire (1,289 copies) from the Taiwan Database of Children and Youth in Poverty of the Taiwan Fund for Children and Families carried out to survey the daily life activities of junior school students from underprivileged families is applied to conduct experiments by using data mining techniques to predict the key factors that influence the academic performance of disadvantaged teenagers. At the same time, Naïve Bayes, JRip and Decision Tree classification algorithms are adopted to construct prediction models and assess the performance of the models.
The outcome of this study shows the level of willingness to receive education, the level of comprehension of courses taught in school, and the level of effort invested in studying are the key factors that influence academic performance. Meanwhile, the JRip prediction model has the best performance.
