利用特徵選取與增加樣本來提高決策樹在信用卡預測的準確度

Using Feature Selection and Addition Samples Improve the Accuracy of the Decision Tree in Credit Card Forecasting

電腦及網路的發達成長的速度越來越快，加上網路科技的發展，使資訊的取得和收集變更便利，因此除了傳統的統計分析工具外，各種用來分析大量資料及許多變數的技術──決策樹演算法也逐漸受到重視。本研究利用（1）特徵選取變數及（2）增加新樣本數來研究是否對於決策樹演算法的準確度有影響，其資料庫是以國內某家銀行信用卡資料庫為樣本做增加樣本及特徵選取變數來預測準確度之變化。研究結果顯示，無論是特徵選取變數或是增加新樣本數都可以提高決策樹的準確度，而特徵選取變數方面所得到的準確度比增加樣本數來的高。

Data and Information can get very easily in the internet epoch. But the traditional statistical analysis tools are limited to analysis the high dimensionality data, the decision tree algorithm is popular which is the one tool of the data mining classification techniques. In this study, we use the feature selection variables and increase the number of new samples to study whether the decision tree algorithm for the accuracy of the influences, the database comes from the department of credit card of a domestic bank corporation. The results show that both the feature selection variables and the addition new samples increase the accuracy of the decision tree, and the accuracy of the feature selection variables is higher than that of the addition new samples.