層級貝式購物籃分析模型之研究

Abstract

企業必須在快速變化的市場裡要開發有價值的顧客和維持有利潤的顧客關係。而如何找到對的客戶群來進行正確的產品行銷在行銷領域裡是個值得研究的主題，但傳統方法裡以人口變數為基礎的方法並不能有效的解決客戶區隔的問題；因此，企業如何能將龐大的客戶人數進行有效率且正確的群體區隔，進而針對不同的顧客群來做行銷活動，在這激烈的市場競爭中愈形重要。對於不同的客戶進行異質、動態且準確的行銷策略，能夠減少行銷預算浪費並增進執行效率。本研究擬利用馬可夫鏈及層級貝式統計建立顧客購買移轉機率矩陣與購買機率之後驗參數估計，並依照購物組合將顧客投入集群分析來進行市場區隔，依照購買移轉機率矩陣，來對不同區隔的客戶做不同的產品推薦；希冀能找到跳脫傳統思維為基礎的客戶區隔方法，讓企業在進行行銷活動時，能夠更有效地針對正確的客戶來做產品推薦。

本研究所提出之計算模式，是應用了層級貝氏統計(Bayes Model)、馬可夫鏈(Markov Chain)，來建立顧客購買機率矩陣，利用RFM模型描述顧客之歷史購買紀錄。貝氏統計方法可幫助我們推估並導入每個顧客的後驗分配，馬可夫鏈則是用來模擬推算顧客在每期購買狀態改變的機率，結合此兩種方法來估計每位顧客的產品購買移轉機率情況，並予之實行產品推薦策略。本研究以國內某3c業者的客戶資料庫進行實証，利用本模型的計算方式相對其他計算方法與實際資料加以比較。

The enterprises have to develop the valuable customer relationship and keep the advantages in this fast changing market. Digging the profit out of the customer relation is an important matter. Therefore, how to select the correct customer segment in the marketing area is the subject worth studying. But the traditional method takes the population variables as the segmenting foundation which does certainly not to be the effective solution. Thus, the enterprise must improve the efficiency in finding the right customer community of huge customer population, then applying marketing strategy to the different customer groups, and keep analyzing the customer data and changing the marking plans in this intense competing market. Holding the dynamical and heterogenic strategy is also a good way to save enterprise budget and improve efficiency. This thesis use Bays statistics model and Markov Chain to build up the migration matrix. Then I use RFM model to define the purchase statement of the customers and Hierarchical Bayes Methodology to compute postier distribution. This system is to offer a commendation reference for marketing managers.