資料探勘(Data Mining)是一門整合不同領域的研究方法，包含資料庫方法、機器學習、統計以及人工智慧等領域。運用此技術，能從大量資料中找出隱藏的知識，來幫助企業歸納與分析資料，並做為判斷與決策之輔助工具。 本研究主要目的是運用資料探勘技術建構國小高年級學生學業成就之預測模式，研究過程中以數學和國語學業成就為輸出變數，以SQL Server 2008為建立預測模型之工具，使用決策樹、類神經網路以及線性迴歸等演算法建立預測模型，在進行變數篩選時，亦利用SPSS進行輸入變數的篩選，以比較不同的篩選方式所得到的結果。評估各個預測模型的預測結果後，發現三種演算法所建立的預測模型，在相同的變數篩選方式下，預測結果並無顯著差異，影響預測結果的關鍵在於輸入變數的篩選方式。 學業成就預測模型可預測學生未來學業成就發展趨勢，學生家長及老師可依據預測模型的結果，及早採取因應措施；學校管理階層進行教學規畫及學生編班時，可依據預測模型的結果，進行適當的調配，提高學生學業成就。Data mining integrate different fields of study, including data warehousing, machine learning, statistics and artificial intelligence. Using this technique, the hidden knowledge could be found from large amounts of data, and help companies to summarize and analyze data. The results of data analysis can be helpful when making decisions. In this study, we build the predictive model of academic achievement for senior students in elementary schools by Microsoft SQL Server 2008. The algorithms for the predictive model contain decision trees, artificial neural networks and linear regression. In order to compare the forecasted results, this study adopts Microsoft SQL Server 2008 and SPSS Statistics 17.0 to find out input variables. According to the results, we find out no matter what algorithm is used, the forecasted results do not have significant difference. The key factors that influence the forecasted results are the method of finding out input variables. The predictive model can predict the stream of academic achievement for senior students in elementary schools. The parents of the students or the teachers can do some promotion activities according to the forecasted results. The forecasted results also serve as references for school managers or administrators while making decisions in their favor.
