資料採礦模式於學校整併指標之應用與評估

Application and Assessment of Data Mining Models in School Consolidation Indicators

因應少子女化的衝擊，小型學校進行整併或裁撤已是必要策略之一，教育部遂於 2006年2月14日提出小型學校發展評估指標，供各縣市政府參考運用。在運用指標進行分析時，若能有大數據的思維，並發展適切的資料採礦模式，將有助於各縣市政府進行學校整併。本研究的研究目的即探討如何基於大數據思維整合不同資料庫，將資料採礦技術運用於教育統計資料中，以利學校整併工作的執行。本研究依據教育部小型學校發展評估指標，整合現行不同資料庫針對個案縣市轄區內所有國民小學進行相關資料蒐集。本研究所運用的資料採礦模式有分類與迴歸樹、類神經網路、決策樹、支援向量機、貝氏網路等五種，研究結果發現五種模式具有正確率高與便於解讀的優點。依據研究結果，本研究提出學校整併應整合教育。

Because of tendency of declining birthrate, it is seen as necessary to consolidate or abolish the small schools. The Ministry of Education then provided “Small School Development Evaluation Indicators” to county and city governments in February 2006. In depth analysis of the indicator data based on Big Data to develop data mining analysis model and operational definition of each indicator, is helpful for county and city governments consolidating small schools. This article aims to study how to integrate different databases based on Big Data thinking, and use data mining methods in education statistics, to facilitate school consolidation. According to the Ministry of Education indicators, this article integrated governance databases to collect the related data of all elementary schools. This article used supervised models, including Classification and Regression Tree, Neural Network, Decision Tree, Support Vector Machine, and Bayesian Network. The results reveal that five models have higher correction rate and are easy to read. According to the results, when consolidating small schools, education, population and geographic databases should be integrated. Besides, empirical data assessment and supervision should be adopted. The governance institution and each school can adopt operational definition of each indicator to calculate the relative position.