一直以來企業為了提高商品銷售量通常利用關聯規則探勘技術從交易資料庫中尋找頻繁項目集，判斷「何種商品」經常被一起購買，當該商品組合經常被消費者購買，則表示此商品組合能有效提升消費者之購買意圖。然而，隨著極端酷熱之天數越出頻繁，近幾年來極端氣候對人類生活之影響不斷地嚴重加劇，中央氣象局指出臺灣近年來高溫日數相較於往年大幅增加一倍之多。因此過去方法挖掘之商品組合能否適用於所有氣溫狀況，而極端高溫之條件下是否會出現不同之商品組合即為一當前值得研究之議題。 為了解決上述問題，本研究提出一套新方法，嘗試從階層頻繁項目集中找到酷熱條件下特定出現之商品組合。延續過去關聯規則之概念，有別於以往利用全年交易資料進行關聯分析，本研究加入氣溫作為區間加以分析，希望找出相較於過去研究更為符合酷熱氣溫下之品項購買組合。本研究將顧客交易資料分成高溫、非高溫兩區段，透過改良之APRIORI演算法應用階層關聯規則並藉由獨立樣本T檢定之結果作為篩選條件，保留頻繁項目集中至少包含一銷售量對氣溫具顯著差異商品之組合，最後扣除兩氣溫區間頻繁項目集中重複出現之品項及頻繁項目集中屬相同分類之組合，即可得到足以代表酷熱條件下特定會出現之品項購買組合。 本研究透過上述方法得到15組頻繁項目組合。當得知未來數日將處於連續高溫時，顧客將有很大機會出現本研究結果之品項組合購買習慣。對於零售商而言，即可依據該購買習慣事先制定相應之商品組合促銷等行銷策略，提高顧客來店之購買意願，並達到企業欲提升商品銷售量之目的。 ;Retailers usually use the Association Rules to find Frequent Itemsets from the transaction database to determine “what products” are often purchased together. However, as the number of extreme hot days increase dramatically, it is important to understand what product combinations are popular during the extreme weather.
Even though some frequent pattern discovering algorithm were designed to find weather related patterns. None of the taking extreme weather into consideration. This study divides customer transaction data into two data sets, High-Temperature and Non-High-Temperature and applies Multiple-Level Association Rules to design an improved algorithm which guarantees that at least one item in the patterns are effected by extreme weathers.
With data coming from a midsize local supermarket, 15 product combinations were identified. The 15 product combinations were purchased frequently in extremely hot days. This finding is useful for making a marketing plan to promote these product combinations in advance.
