在資料探勘的領域中，關聯法則可以反應出顧客購買時對其他物品也需同時購買的需求，現實中除了這樣的消費行為，若遇到商品缺貨或價格調整時，可能會購買替代商品來取代原始商品，因此商品間還存在另外一種關係─替代關係，如同關聯法則一樣，替代關係亦可以提供決策者更多的決策資訊。 本研究以資料探勘為學理基礎，加入經濟學上替代品的觀點發展替代法則的探勘，透過找出銷售交易資料庫中彼此符合經濟學替代品定義的商品來決定其是否具有替代關係，為了處理現實生活中的某些現象，並發展了 、 、 和 四個參數用以調整這些偏差。 以實際的超市交易資料實驗結果顯示，同一類別下的商品其替代法則會與現實生活中直覺較相似，除了完全替代關係，亦能適當地發掘其他部分替代關係，在不同分店和不同時期的消費者替代法則結果是不相同的，顯示地理位置和季節的不同會有不同的購買決策，其中季節因素又比地理位置因素來的重要。 In data mining, association rule is used to discover products that co-purchase frequently within customer purchasing transactions. But if product is out of stock or price adjustment, a customer may be to replace the original purchase of some items with that of others. Therefore it exist another relation, substitution rules, in products. The mining of substitution rules, the same as that of association rules, will get very valuable knowledge in decision support. In this thesis, the mining algorithm for substitution rules is based on the concept of data mining and definition of substitute in the economics. The substitution rules mining algorithm is designed to find out the products that comply with the definition of substitute in the transaction database with four parameter, , , and , for fitting in with real-life. The substitution rules mining algorithm is implemented and experimented against the real supermarket database. It is shown that the algorithm not only can find out the perfect substitute but also can discover some imperfect substitutes. It is also verify that different branch store and period have distinct substitution rules, and season is more important than position.
