本研究主要基於考慮各品項在資料庫中的出現期間，並且在多門檻的環境之下挖掘跨概念階層中，交易之間的關聯規則。 在交易間之關聯規則的研究之中，FITI是基於Apriori演算法所發展出來的方法，在挖掘的過程之中，讀取資料庫的次數過多，而降低了挖掘的效率，而在本研究主要基於FP-tree演算法發展出的方法ITCL_FP-tree(MIS)，挖掘在跨概念階層中交易間的關聯規則使挖掘出來的規則更為詳細，也能更增進挖掘過程的效率，並且考慮到在多門檻限制的研究中所沒有考慮到的，也就是資料庫中品項的出現週期。品項的出現週期主要考慮到該品項並非在整個資料庫中都會出現，可能只會在某一定時間出現，因此我們以建立各品項的出現週期，配合多門檻限制進行挖掘，並且在挖掘關聯規則時將會以各品項的出現週期為基礎，以挖掘更為詳細的關聯規則。 我們利用實際的資料進行實驗驗證，ITCL_FP-tree(MIS)比CL_FP-tree(MIS)以及IT_FP-tree(MIS)能夠採擷更多的關聯規則，包含有跨概念階層的資訊以及不同交易之間的關聯規則，同時，利用了移除多餘的父階品項更可有效的降低多餘的關聯規則數量以及在挖掘過程中的效率。 In our research, we mainly consider the Appearance period of each item in the database and minimum item support (MIS) to mine inter-transactional association rules. In the previous researches of inter-transactional association rules, the methodology they proposed are based on Apriori Algorithm and didn’t with the consideration of product concept hierarchy. The efficiency of their methods is reduced by generating too many candidates for frequent items and the rules wouldn’t give us more detailed information. In our research, we proposed an methodology named ITCL_FP-tree(MIS) for mining inter-transactional association rules not only about items in the atomic level but also cross the concept levels. Considering of each item’s natures of appearance, we count the support of each item with the consideration of the Appearance period of items instead of count the support by the length of the database. This will solve the rare item problem. In our experiment, we use the real-life data for verifying ITCL_FP-tree(MIS) can mine more frequent rules than CL_FP-tree(MIS) and IT_FP-tree(MIS) which we proposed for mining inter-transactional association rules among items in the atomic level of concept hierarchy. And we also use the concept of gap for pruning the frequent but redundant parent items.
