本研究探討如何找出消費者購買行為之間的關聯，透過最小生成樹演算法而不是以往的關聯法則演算法等，以往的關聯法則演算法需透過決策者自行定義門檻值去挖掘出關聯法則，門檻值沒有準確的定值容易找出過量的關聯法則使得決策者無從觀察，且關聯法則忽略了消費者購物的特性，忽略了產品購買數量與金額的事實，因此本研究使用最小生成樹在交易資料中找出兩兩最具相關的產品種類，透過相互資訊進行產品與產品之間的顯著性檢測證明其在統計上是顯著相關的，並利用中心程度來去找出顧客的購物籃中最常被購買出現在各個購物清單的關鍵產品，最後使用 utility scores 去算出最有價值的關聯法則供給零售業經理人進行促銷活動等。;This study explores how to find the correlation between consumers′ purchasing behaviors, through the minimum spanning tree algorithm instead of the previous association rule mining. The previous association rule algorithm needs to define the threshold by the decision maker to mine the association rules , however how to define the threshold is a difficult question, it is easy to find excessive useless association rules that make it impossible for decision makers to observe, and the association rule ignores the characteristics of consumer shopping eg. the fact that the number and amount of product purchase. Therefore, this study uses the minimum spanning tree in identify the most relevant product category in the transaction database, and use the mutual information to conduct product-to-product significant test to prove that they are statistically significantly related, and use the degree centrality to find the key product in the customer′s shopping basket. They are often purchased for key products that appear on various shopping lists, and finally use utility scores to calculate the most valuable association rules for retail managers to carry out promotional activities.
