在微利的時代中，企業投入大量資源增加對顧客的瞭解並與顧客建立良好關係，並界定不同價值的顧客群。企業以不同的產品、不同的通路滿足不同區隔顧客的個別需求，並在關鍵時刻，持續的與不同層次的顧客溝通，強化顧客的價值貢獻，希望提高顧客滿意度(Customer Satisfaction)及顧客忠誠度(Customer Loyalty)。本研究從顧客對公司的貢獻價值方向，提出以RFMP分析模型的「最近一次的購買日期」(R)、「一段期間內的購買頻率」(F)、「一段期間內的購買金額」(M)與「一段期間內的促銷產品購買頻率」(P)為基礎，以顧客歷史交易資料進行顧客價值分析，應用倒傳遞類神經網路技術於個別顧客的購買歷史記錄進行分析，針對個別顧客的購買行為進行預測，建立顧客價值指標做為顧客價值的預測模型 。最後利用發展的FN_DBSCAN群集的演算法找出一群有數量限制的高價值顧客群。 In tiny profit years, enterprises are investing many resources to increase understanding of customers and establish good relationship with customers, and mark off different value customers. Enterprises use different products, different place to satisfied the different needs in separate customers. In a critical moment, keep on communicating with different level customers, to strengthen the contribution of customers, in a hope to raise customer satisfaction and customer royalty. In this thesis, based on the RFMP(Recency, Frequency, Monetary, Promotion) analysis model, we use customer history transaction data to perform customer value analysis. Apply back-propagating neural networks technology to analyse the individual customer history transaction data and forecast the consume behavior, to establish the customer value index as customer value prediction model. Finally, we use the FN_DBSCAN algorithm that developed in the thesis to find the fixed number of prior customers.
