A Discount-Based Approach for Collaborative Filtering Recommendation

Due to the rapid development of E-commerce, the number of online stores and that of products have been increasing drastically. Two issues have been raised in this regard: customers waste too much time browsing un-needed products, and the even more competition among online stores happens. Personalized recommendation systems (PRS) are developed and considered as a solution to the first issue. PRS recommend customers products according to their interests, aiming to develop potential consumers, increase purchasing rates and to enhance customer royalty. To face the competition, online stores provide diversified promotion strategies. Yet, the current approaches to product recommendation are only based on consumers'preferences, rather than taking the discount of sale price into consideration, thus leading to insufficient incentives for customers to purchase. The aim of this study is to prove that discounts really have impacts on the consumer purchase intention in online-purchasing, and that they have further effect on the accuracy of PRS. To fulfill this research aim, this study explicitly focuses on the cosmetic online-stores which sell mostly similar products and frequently launch promotions. It uses the price-pack deal and products prices as variables to exam the effects of discounts on the customer purchase intension. This study also proposes the purchase-intension decision-tree and formula to incorporate into the collaborative filtering approach. A discount-based collaborative filtering (DB-CF) recommendation approach is thus established with benefits of aligning PRS with retailers' promotion strategies, increasing the exposures of promoted products, facilitating customers' decision making processes and improving the accuracy of PRS.近幾年來由於電子商務快速發展，網路商店及商品數量大量增加，進而導致兩個問題：一、顧客浪費許多時間在瀏覽無用的商品上；二、隨著網路商店大量增加，商家間的競爭也越發激烈。為了解決第一個問題，個人化推薦系統因應而生，主要是根據顧客的興趣來推薦商品，目的為挖掘潛在客戶，推薦其他商品或提高顧客忠誠度。在激烈競爭下，商家時常需要推出各式各樣的促銷策略來吸引顧客，其目的與推薦系統十分相似。 然而，現有的推薦方法只考慮到消費者的偏好，並未將商品的折扣因素考慮進去，造成沒有足夠的誘因促使顧客購買推薦的商品、商家想要推銷的產品不一定會被列入推薦清單以及顧客沒有發現自己需要的東西正在促銷等問題。 過去研究指出，折扣呈現型式和折扣幅度是影響消費者購買行為的重要因素。為了證明折扣資訊確實會影響網路拍賣中消費者的購買意願，進而影響推薦系統的準確度，本研究以大多販賣相同商品且時常推出促銷活動的美妝商店為例，以折扣條件及商品價格為變數，分析折扣資訊對消費者購買意願的影響。 最後，本研究提出購買意願決策樹以及公式，和協同過濾式推薦方法做結合，提出一個以折扣資訊為基礎的協同過濾式(Discount-Based Collaborative Filtering,DB-CF)推薦方法。我們的推薦方法讓推薦系統能夠配合商家的促銷策略，增加促銷商品的曝光率，輔助消費者進行產品選擇，並增進推薦系統的準確度。摘要 i ABSTRACT ii 目錄 iii 表目錄 v 圖目錄 vi 第1章 緒論 1 1.1 研究背景及動機 1 1.2 研究目的 2 1.3 主要貢獻 2 1.4 論文架構 3 第2章 文獻探討 4 2.1 促銷活動與消費者購買意願 4 2.1.1 折扣呈現型式對購買意願的影響 4 2.1.2 折扣幅度對購買意願的影響 5 2.2 推薦系統 5 2.2.1 內容導向式推薦方法Content-based Recommendation 6 2.2.2 協同過濾式推薦方法Collaborative FilteringRecommendation 6 2.2.3 混合式推薦方法Hybrid Recommendation 8 第3章 研究方法 9 3.1 問題描述與基本概念 9 3.2 折扣資訊分析方法 10 3.2.1 折扣影響購買意願公式與購買率定義 10 3.2.2 折扣呈現型式分析方法 11 3.2.3 折扣幅度分析方法 12 3.3 結合折扣資訊的協同式過濾式推薦方法DB-CF 12 第4章 折扣資訊的分析 14 4.1 折扣呈現型式分析 14 4.1.1 折扣呈現型式 14 4.1.2 以折扣條件為變數 22 4.1.3 折扣呈現型式決策樹 27 4.2 折扣幅度分析 31 4.2.1 折扣幅度 31 4.2.2 以商品價格為變數 33 4.2.3 折扣幅度影響購買意願公式定義 45 4.3 折扣資訊分析結論 47 第5章 實驗與結果分析 48 5.1 協同過濾式推薦方法實驗結果 48 5.2 DB-CF推薦方法實驗結果 50 5.2.1 DB-CF推薦方法實驗一 50 5.2.2 DB-CF方法實驗二 51 5.2.3 DB-CF推薦方法實驗三 55 5.2.4 DB-CF推薦方法實驗四 57 第6章 結論與未來研究方向 61 參考文獻 6