隨著資訊科技的日新月異,電子商務營運模式已成為實體企業的殺手級應用,故在虛擬的交易平台裡,企業為了彌補無法面對面互動交易所造成的資訊缺乏,必須盡可能的搜集並分析顧客所留下來的歷史交易資訊;此外電子商務營業額以及商品銷售量的蓬勃與否與商品本身的性質有絕對的相關,假設商品屬於鑑賞性的商品而並非消耗性的商品,則消費週期便會拉長、重覆購買的現象也會減少,因此整體的消費頻率以及購買數量便會降低,使得單純的敘述統計分析無法有效地反映少量交易資料背後所隱含的決策性意義。 本研究針對B2C之電子商務網站,利用層級貝氏模型的分析方法,建構會員的異質性消費模型。相關的顧客資訊,包含了RFM模型中的購買頻率(Frequency)以及購買金額(Monetary)、會員本身的人口統計變數資訊。在分析的方法方面,運用馬可夫鏈蒙地卡羅法(Markov Chain Monte Carlo, MCMC)技術進行模式中參數係數的估計;不同於以往關於消費模式建構的研究假設購買頻率與購買金額為相互獨立,本研究將購買頻率與購買金額假設為相依變數,透過層級貝氏法搭配MCMC技術來了解顧客消費偏好,本研究之方法除了可增加模型解釋維度的彈性之外,亦可進行延伸性的行銷分析,如加入異質性的分群機制、或品牌選擇模型等,來強化整體消費模式的參考價值,達到一對一行銷的目的。
E-Commerce business model has already become the killer application for enterprise as IT keeps improving. Enterprise must try to search and analyze customer’s historical trading information to deal with the lack of communication in virtual trading platform. However, there are strong relation between business volume and attributes of merchandise. If one product do not belong to consumable and has longer purchase cycle, this product’s purchase frequency and purchase quantity will also decline. Using description statistic analyze method can not reflect the decision meaning behind such less trading data. This research constructs a consuming model with heterogeneity by Hierarchical Bayesian Model to help enterprise using customer’s information in Data Base which reveals longer purchasing frequency and lesser purchasing quantity. This model not only connects purchasing information but also member’s demographic statistic data. And apply Markov Chain Monte Carlo (MCMC) simulation technique to estimate the coefficient to represent customer’s preference. It is easy to practice other marketing analysis such as clustering and brand choosing model to enhance the reference value of this model, and reaching the goal of one-to-one marketing