顧客關係管理關係著一個組織的興衰,在學術上及實務上都是一個重要的研究議題,且由於資訊科技的進步及顧客關係管理的資訊化,也更引發了各種新的相關研究議題。而關於維持顧客關係的議題中,要如何能精準鎖定顧客流向、降低顧客流失以及降低浪費的成本一直都是組織憂慮的問題。 本研究利用目前最常用來分析顧客價值的測量方式,包含:RFM、顧客趨勢分析,以及顧客個人特徵等資訊做為研究變數,並結合支援向量迴歸(Support Vector Regression, SVR),建立一個可提供醫美產業之預測顧客流失之模型。其中,本研究對於個人特徵變數上,特別加入了『星座因素』、『年齡因素』、『血型因素』以及『性別因素』做為區別顧客的特徵。關於支援向量迴歸(SVR)的應用,傳統的常將之應用於圖像辨識上,而近年來常被使用在商業模型的預測,如股票、大盤指數之預測,而本研究係利用SVR的預測彈性,來萃取目前醫美產業的顧客特徵,進而訓練出一個有效的顧客流失之預測模型。由資料預測的結果顯示,不同的星座對於顧客消費行為是具有一定的影響,且不同星座的消費態度確實會影響顧客對於產品的態度;另外,在個人年齡因素上,年齡因素也有效的區別出顧客間的消費能力以及對於醫美產品的喜好程度。然而,星座的影響力對於消費行為上不全然在所有情況都符合,星座因素必須在沒有『風險成本因素』的影響下,才能完全顯示出來。最後,本研究也透過『預測後資料間仍保有的顧客屬性』,來間接地證明上述的預測模型,是具有一定的準確性及可行性。 本研究透過過去的交易資料,以及相關的參數設定,訓練出一個符合目前醫美產業的預測顧客流失模型,且經由誤差驗證以及資料分析後,顯示本研究所提出的模型是具有一定的預測能力,此外同時也印證了,SVR除了應用於傳統的圖像辨識上,在顧客流失的預測上,SVR一樣具有良好的預測效果。
Customer Relationship Management (CRM) is a very active research area for decades, and has accumulated voluminous literature. For many organizations, it would be very difficult to analyze precisely all the customer attrition, especially in medical cosmetology in Taiwan. We, therefore, utilized several kinds of transaction information which are frequently used in evaluating customer value as our variables. These variables are RFM, Customer Trend Value and demographic variables. To combine the measurement of demographics with the psychological insights of psychographics, we use astrology as a variable for segmentation purposes. The main objective of this study is intended to use all the variables to establish a Support Vector Regression (SVR) model and tries to provide some suggestions to the medical cosmetology industry. Recently, SVR has been used to solve regression and prediction problems. In this study, we apply the SVR to customer attrition forecasting. In particular, the transaction data are usually complex. To ensure all the transaction date are correct, we use SQL command to check all the data prudently so the risk of incorrectly calculating the SVR model is greatly reduced. Our experimental results show that astrology variables have a significant impact on attitude toward cosmetic product if there is no financial risk. Finally, we can conclude that the customer attrition model based on SVR obtains a good balance among fitting ability, generalization ability and model stability by the Mean Square Error (MSE) test and discussion with experts.