隨著數位化時代的來臨，資料量暴增導致資訊過載，對於日理萬機的高階主管，要在短時間內消化大量資料，並且在對的時間，給予對的行銷方案實屬難事，為了解決資訊過載的問題，本研究提出了摘要化交易資料庫的演算法，在眾多資料中找出最具有代表性的交易資料，以減少資訊閱讀的時間。期望協助高階主管進行快速決策，讓高階主管可以使用少數具有高度可讀性的代表資料，來窺探整體線上交易零售資料庫，以快速得知整體的銷售概況。 本篇研究使用K-medoids、Balanced K-means以及Genetic Algorithm演算法運算，找出最能代表線上交易零售資料庫的交易紀錄，並且比較三者的總成本，而總成本是由代表成本及代表不平均成本組成，最後期望以Genetic Algorithm，來改善使用K-medoids運算時的代表問題，在降低代表成本的同時，也提高代表性。;With the digital generation coming, the data has been explosive growth and causes the information overloading. For a senior manager, it is hard to digest so much data and make a right marketing decision in right time. In order to resolve the problem of information overloading, this research provide an algorithm of transaction data reduction. It can reduce the time of searching the information by discovering the most representative data from the large data set. We expect to help senior managers to make the decision more efficiently. With making good use of those representative data, they can see whole the online transaction retail database and realize the basic facts of all the sales in the short time. This research will adopt the K-medoids, Balanced K-means and Genetic Algorithm to discover the most representative transaction data from the online transaction retail database. We will also compare the total cost of the three algorithm which is composed of representative cost and representative imbalanced cost. We propose the Genetic Algorithm can improve the representative problem, which is able to reduce the representative cost and also improve the representative of the data.
