以機器學習技術建構顧客回購率預測模型：以某手工皂原料電子商務網站為例;Building a customer repurchase rate prediction model with machine learning technology: Taking a handmade soap raw material e-commerce website as an example

NCUIR > School of Management at National Central University > Executive Master of Information Management > Electronic Thesis & Dissertation > Item 987654321/88346

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/88346

Title:	以機器學習技術建構顧客回購率預測模型：以某手工皂原料電子商務網站為例;Building a customer repurchase rate prediction model with machine learning technology: Taking a handmade soap raw material e-commerce website as an example
Authors:	曾建坤;曾建坤
Contributors:	資訊管理學系在職專班
Keywords:	機器學習;資料探勘;回購率;電子商務
Date:	2022-05-17
Issue Date:	2022-07-13 23:41:55 (UTC+8)
Publisher:	國立中央大學
Abstract:	隨著全球電子商務網站購物人數的增長，全球各地的企業販賣商品均嘗試將資源集中在電子商務網站上來提高其競爭優勢，在台灣，手工皂產業如雨後春筍般蓬勃發展，每年創造數億的商機。因此，面對如此龐大的商機，提供手工皂的原料市場競爭亦非常激烈，研究個案已有經營多年業績良好的實體店面，畢竟實體傳統零售店面多有地域性限制，客戶多為區域型客戶，必須運用網路跨越地域，擴大市場接觸面，才能掌握網路無遠弗屆、一觸即發的商機。研究個案把行銷主要資源放在電子商務市場，經營網站商機又以開發網站新會員與舊會員回購商機為兩大主力。在市場上可以多重選擇購買廠商的情況下，客戶隨時會改變購買行為，所以行銷界認為獲得一個新客戶的成本是保留一個老客戶成本的5~6倍。因此，增加銷售資源在曾經購買的客戶群發掘更多的商業機會，客戶不斷的回頭購買產品，擁有越來越多忠實會員客戶，無可置喙的就會讓我們的銷售工作變得容易，業績穩定成長。透過文獻探討所羅列的研究中，整理出包含：顧客關係管理，網站與產品到訪點擊流量，與客戶的關係長度，購買次數、購買金額，Electronic Direct Mail (EDM)行銷收信、開信次數，購買紀錄，對特價產品偏好，退換貨因素，景氣與失業率等因素，皆被視為與電子商務回購有關的網站行為相關因子。透過機器學習的技術，尤其隨機森林、羅吉斯回歸、梯度提升機建立預測回購模型，AUC的表現均超過0.7以上。此外，本研究依Orange提供的資訊增益比技術，針對不同研究變數對目標變數所能提供的資訊值，來判斷其對預測回購模型的重要性排序，其前五名排序為1. 特價產品購買次數2. 寄送電子行銷信件打開信件次數3.加入會員以後購買產品次數4.會員最近三個月點閱網站首頁次數5.會員下訂單後取消產品次數。透過本研究掌握顧客最佳網站行為變數，構思出增加回購率的產品行銷方案及增加回購的會員數策略，達到建構預測模型的目的，有效吸引消費者回購，不斷創造佳績。;With the increase in the number of shoppers on e-commerce websites around the world, companies around the world are trying to concentrate resources on e-commerce websites to improve their competitive advantages. In Taiwan, the handmade soap industry is booming like mushrooms after a spring rain, creating hundreds of millions of business opportunities every year. . Therefore, in the face of such huge business opportunities, the market competition for raw materials for handmade soaps is also very fierce. The research case has been a brick and mortar store with good performance for many years. It is necessary to use the Internet to cross regions and expand market contacts in order to grasp the unlimited and imminent business opportunities of the Internet. In the case study, the main resources of marketing are placed in the e-commerce market, and the business opportunities of operating the website are the development of new website members and the repurchase of old members as the two main forces. In the market where there are multiple options for purchasing manufacturers, customers can change their purchasing behavior at any time, so the marketing community believes that the cost of acquiring a new customer is 5-6 times the cost of retaining an old customer. Therefore, increasing sales resources to explore more business opportunities in the customer base that once purchased, customers continue to return to buy products, and have more and more loyal member customers, which will make our sales work easier and performance. stable growth. Through the literature research, the researches listed include: customer relationship management, website and product visits, click traffic, relationship length with customers, number of purchases, purchase amount, Electronic Direct Mail (EDM) marketing receipts, and openings. , purchase history, preference for special products, return factors, prosperity and unemployment, etc., are all considered as factors related to website behavior related to e-commerce repurchase. Through machine learning techniques, especially random forests, logis regression, and gradient boosting machines to establish predictive repurchase models, the performance of AUC is more than 0.7. In addition, according to the information gain ratio technology provided by Orange, this study judges the importance of different research variables to the target variable according to the information value provided by the target variable, and the top five are ranked as 1. Special products Number of purchases 2. Number of emails sent and opened 3. Number of product purchases after joining as a member 4. Number of times a member has clicked on the homepage of the website in the last three months 5. Number of products cancelled after a member placed an order. Through this research, we grasp the best website behavior variables of customers, devise a product marketing plan to increase the repurchase rate and a strategy to increase the number of members who repurchase, achieve the purpose of constructing a prediction model, effectively attract consumers to repurchase, and continuously create good results.
Appears in Collections:	[Executive Master of Information Management] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	81	View/Open

社群 sharing

Loading...