資料探勘應用之研究:零售業的RFM分析架構

Abstract

在資料庫行銷領域中，RFM模型一直是一個很重要的角色，他能提供一個簡單的框架去量化顧客。隨著時代的演進，RFM模型結合資料採礦能使企業對於顧客的分析更透徹，不論是用於對顧客進行分群或是分析顧客價值。本研究使用公開平台的交易資料進行分析，以真實零售商之交易資料分析該企業的顧客，嘗試以RFM模型結合資料採礦的方法，對客戶進行分群，最後建立預測模型並驗證其預測力，同時本研究也著重在資料前處理的描寫。本研究以二階段集群分析結合RFM指標將顧客分成四群，並且將分群後的結果作為目標變數，以決策樹分析與判別分析建立預測模型，最後發現判別分析之預測率較好，但決策樹擁有較易解釋的規則。In the field of database marketing, the Recency, Frequency, Monetary model has always played an important role, it provides a simple framework to quantify customers. With the evolution of the technology, the RFM model combined with data mining enables companies to analyze customers more thoroughly, whether it is used to segment customers or analyze customer value. This study uses the transaction data of the open data platform, and analyzes the customers of the retailer's transaction data. It attempts to combine the data mining method with the RFM model, and then builds the predicting model and verifies its predictability. This study also focuses on the process of data pre-processing. In this study, the two-phase cluster analysis combined with the RFM index divides the customers into four groups, and the results of the grouping are used as the target variables. The prediction model is established by decision tree analysis and discriminant analysis. Finally, the prediction rate of the discriminant analysis is better, but the decision tree is easier to explain.