應用決策樹分析於資料探勘之研究－以預測混凝土抗壓強度為例

The Study of Applying Decision Tree Approach to Data Mining - An Example of Forecast Compressive Strength of Concrete

王來旺(Lai-Wang Wang) ； 呂宜倫(Yi-Lun Lu)

《商業現代化學刊》 8卷2期 (2016/10) Pp. 115-130

https://doi.org/10.6132/JCM.2016.8.2.07

摘要

資料探勘(Data Mining)技術，是近年被廣泛應用於各領域來協助分析大量資料的技術，主要目的是希望透過資料探勘技術，從大量的歷史資料中挖掘出其中附有價值的知識或訊息。相較於類神經網路，決策樹模型擁有較好的理解性，在解釋上也較容易，所以本研究先使用皮爾森相關係數篩選可用之變數，再使用決策樹中的分類與迴歸樹(Classification And Regression Tree)以及卡方自動交互檢視法(Chi-squared Automatic Interaction Detection)進行高性能混凝土之抗壓強度預測模型的建立進而使用衡量預測誤差指標(MAD、MSE、MAPE、SMAPE、RMSPE)比較，研究結果得知，分類與迴歸樹(Classification And Regression Tree)模型在高性能混凝土之抗壓強度預測較準確。

關鍵字

資料探勘 ； 決策樹 ； 皮爾森相關係數 ； 衡量預測誤差指標

並列摘要

Data mining technology, has been widely used in various fields to analyze a large amount of data, the main purpose is to through the data mining technology, which with valuable knowledge or information mining from a large amount of historical data. Compared with neural networks, decision tree model has a better understanding of in the interpretation, is easy, so this study first use Pearson correlation coefficient screening available variables, and then use the classification and regression tree in decision tree and chi square automatic interaction detection method to establish high performance concrete compressive strength prediction model and using a measure of forecast error index (MAD, MSE, MAPE, SMAPE, RMSPE), the research results show that the classification and regression tree in the model The prediction of high performance concrete compressive strength accurately.

並列關鍵字

Data mining ； decision tree ； Pearson correlation coefficient ； Predictive error index