應用資料探勘技術於RFM顧客價值區隔之研究

Applying Data Mining Technology to RFM Customer Value Segmentation

王智立(ChihLi Wang) ； 郭若宣(JoHsuan Kuo)

《智慧科技與應用統計學報》 19卷1期 (2021/08) Pp. 36-64

摘要

在信用卡發展趨於穩定的市場，金融機構能瞭解信用卡用戶之顧客價值與消費行為是至關重要的研究課題。以往資料庫行銷中利用RFM模型來分析顧客價值，建立簡單架構分析顧客之消費行為，將顧客進行評分。但傳統RFM模型使用的等間距方法進行分群，而本研究使用傳統方法分群後，將數據以各群平均值取代原始數據，並使用兩階段集群分析將顧客分群，以瞭解不同顧客類型及其價值。隨著大數據之發展，資料探勘被廣泛應用，演算技術也越來越多種。本研究利用 RFM模型中之三項指標進行顧客價值之兩階段集群分析，再利用顧客基本資料使用C5.0決策樹、隨機森林與支援向量機等分類方法將顧客進行分析，建立顧客類型的預測模型，以便後續能直接判斷其顧客類型。

關鍵字

RFM模型 ； 資料探勘 ； 兩階段集群分析 ； 分類分析

並列摘要

In a market where credit card development is stabilizing. It is important research topics for financial institutions to under stand the customer value and consumption behavior of credit card users. In the past, database marketing used RFM models to analyze customer value. Build a simple structure to quantify customers and grade customers. However, traditional RFM models use an equally spaced method to group customers. This research uses traditional methods to group the data. Replace the original data with the average of each group, and use two-stage cluster analysis to group customers. To understand different customer types and value. With the development of big data, data mining have been widely used. The algorithms have become more and more diverse. This research will use the original data of the three indicators in the RFM model to conduct a cluster analysis of customer value, and then use customer basic data to analyze customers using C5.0 decision tree, random forest and support vector machine classification methods. Finally building a predictive model of customer type so that the customer type can be directly judged later.

並列關鍵字

RFM model ； Data mining ； Cluster analysis ； Classification analysis