從信用卡交易紀錄探勘消費者衝動性購買行為

Exploring Customer Impulse Buying Behavior from Credit Card Transaction Records

王偉丞(Wen-Cheng Wang)

指導教授 ： 任立中

國立臺灣大學/管理學院/國際企業學研究所/碩士(2015年)

https://doi.org/10.6342/NTU.2015.00060

摘要

衝動性購買行為屬於一種突發的、難以抗拒的消費行為，在過去的研究上，大多是以心理層面的研究出發，探討任何可能的行銷刺激、購物環境、行為變數、人格特質等影響因素。過去傳統的方式，是以問卷分析的方式來解析研究引發衝動性購買行為的影響因子；但由於衝動性購買的突發性，導致研究者常常無法立即性地得知確切的原因究竟是什麼；抑或在測量消費者購買前跟購買後的態度差異時，也不一定能獲得最真實、態度一致的結果，因此導致研究上的限制。
然而衝動性購買在消費者行為上是一種極為普遍的現象，其交易行為幾乎隨時都在發生。從過去的研究中，可以知道當涉及某些特定的商店、特定產品或是產業時，會有衝動性購買行為的比例相對提升許多；且更進一步顯示出，具有衝動性購買行為特質的顧客，對企業主所追求的顧客價值相對來說也更大。因此對於衝動性購買的研究不管在學術或是實務上無疑是一項發展趨勢。
現今資料庫探勘的崛起、資訊科技、網路的普及，將社會帶向了一個全新的面貌。而在大數據時代的來臨，無數的交易行為代表著無數的交易資料被一一紀錄下來，這些紀錄代表的即是消費者最真實也是最確切的交易情況。資料庫統計方法的使用，搭配行銷決策的運用，我們可以嘗試去挖掘消費者背後的行為動機。但民眾的衝動性購買行為在過去的發展中，鮮少有利用消費者的交易紀錄來探勘消費者的消費行為特質。
因此，本研究運用資料統計探勘方式，以消費者的信用卡交易紀錄為主，利用兩階段集群分析法，搭配群集分類的動態穩定性檢驗，來確認利用該分析方式是否能有效地從資料庫紀錄中分類出具有衝動性購買行為特質的顧客。
其結果顯示，經由交易紀錄欄位做變數處理的分類變數指標，有效地將顧客分成衝動性與非衝動性購買顧客，且其區隔穩定性的結果為顯著，代表該分群方法妥當，但本次想要抓取的衝動性購買顧客比例卻過低。因此，考慮單一變數分類指標來做顧客分類，並以交叉表卡方檢定來分析檢驗，其分類結果顯著，且抓取的衝動性購買顧客人數和範圍也有效擴大。
根據人口統計變數和交易項目分析結果，我們可以進一步與相關的企業主合作發展各種顧客關係管理方案、一對一客製化行銷策略、推薦系統等，為企業及市場創造更大的價值。

關鍵字

衝動性購買行為 ； 資料庫行銷 ； 資料探勘 ； 集群分析 ； 兩階段集群分析 ； 動態區隔穩定性 ； 卡方檢定 ； 交叉表分析

並列摘要

Impulse buying behavior is an unexpected and irresistible purchasing behavior. From all the past research on impulse buying, most researchers focused on psychological factors like purchasing environment, personality traits, marketing stimuli and behavior parameters etc. by using traditional questionnaire analysis to study the factors that lead to impulse buying. Because of the characteristic of being unexpected and sudden, researchers couldn’t identify the exact reason immediately. We even couldn’t know the actual outcomes when measuring the difference of purchasing attitudes before buying and after buying. That leads to the restrictions of previous research.
Impulse buying is a very normal behavior of consumer behavior and transactions that could happen anytime and anywhere. According to past research, the percentage of impulse buying behavior would rise when some specific stores and particular industries were involved. Furthermore, the more impulse buying characteristics consumers have, the more customer value enterprises pursue. Therefore, no matter on academic research or on practices, impulse buying researches has already been a developed trend.
The developments of database mining, information technology, and ubiquitous network have changed the society. Countless transaction behavior means countless transaction data are being recorded during the big data era. With the statistical methods on database and the collocation of marketing strategies, we can try to identify the motivations behind the consumer behavior. But there are few research studying impulse buying by using data mining methods on consumer transaction records. Therefore, I tried to use some data mining methods to analyze the credit cards transaction database that targeted customers with impulse buying characteristics by utilizing two-step clustering analysis, demographic variables analysis, and the dynamic segments stability analysis.
According to the first outcome, the variables we chose and dealt with used in two-step clustering analysis from the credit cards database are able to differentiate the impulse buying and non-impulse buying customers. But the percentage of the impulse buying customer was too low. So, I kept tring another method to identify impulse buying customers by using only one classification index that was made from database. The second time’s outcome was siginificant and the number of those impulse buying customers we identified was much more than the first time.
Finally, we can use these results, as well as the demographic variables analysis and the transaction items of industries analysis from records to make customer management programs, one-to-one customized marketing strategies, recommendation systems and to develop the collaboration relationships with related industries and enterprises. With the objective to create bigger and much more value for the market and companies.

並列關鍵字

Impulse Buying ； Database Marketing ； Data Mining ； Cluster Analysis ； Two-Step Cluster Analysis ； Dynamic Segments Stability ； Chi-Square Test ； Cross-Table Analysis

參考文獻

2. 胡鈞傑，2006， "應用貝氏統計方法於建立顧客消費行為預測之時間序列模型"， 臺灣大學國際企業學研究所學位論文 ：1-76。

3. 廖怡芳，2004，“資料庫行銷之應用--以百貨業為例"， 臺灣大學國際企業學研究所學位論文 : 1-122。

5. 彭銘修，2011，"顧客價值與遷徙路徑分析─ 以信用卡資料庫為例" ，臺灣大學國際企業學研究所學位論文 : 1-65。

6. 王嵐，2010，"B2B 企業之顧客採購行為分析-以醫藥公司為例"， 臺灣大學國際企業學研究所學位論文 : 1-63。

7. 劉蕙，2006， "以群集分析與序列型樣輔助資料庫行銷策略"，元智大學資訊管理學系學位論文 : 1-53。