對於各金融機構而言,顧客為企業最珍貴的資產。然而,銀行要如何運用顧客關係管理與行銷手法,將有限的資源發揮最大的效益,那準確的選取潛在顧客則是相當重要的。故本論文目的為利用資料探勘之技術,建置目標顧客預測模型,以找出潛在顧客。 本研究設計為三階段,分別探討資料前置處理、特徵挑選與模型建置。我們針對不平衡資料集使用6種抽樣比例,並利用3種特徵選取法搭配4種分類演算法來建置模型,最後比較其模型效力。 實驗結果顯示,樣本抽樣比例的設定強烈影響分類預測的效果。而分類預測效果最佳之模型是在3:7樣本抽樣比例下,使用R-Square特徵選取法搭配Tree分類演算法。本研究貢獻為建立一套應用於金融業挖掘潛在顧客之模型的研究方法,並可利用預測模型為銀行產出潛在顧客名單,作為行銷決策之參考。
Customers are generally bank and financial institutions’ most vital asset. Thus, it is important for institutions to precisely catch the customers by using limited resource for marking. The purpose of this paper is looking for potential customers by technique of data mining and prediction model. A three-phase study was designed to explore the data pre-processing, feature selection, and model building. We take six samples from imbalance data sets, employed three different methods of feature selection go with four Classification algorithms, and then compare the preference of these models. The results of this experiment showed that the proportion of samples strongly impacts prediction. Under the proportion of 3:7 in sampling, we find the best preference of the model using R-Square to collocate Decision Tree. The paper contributes to bringing the forth method that uses prediction model for bank and financial institutions to look for potential customers.