利用分群化技術發掘消費者適性產品項目

Using Clustering Technology to Find Consumers' Adaptive Products

陳垂呈(Chui-Cheng Chen) ； 余宗憲(Zong-Xian Yu)

《致理學報》 39期 (2019/11) Pp. 245-263

摘要

企業利用資訊技術可以輕易記錄、儲存消費者的交易記錄，面對快速累積大量的數據，若能從中分析消費者興趣傾向，對協助企業提供消費者適性產品行銷策略、及提升市場利基都有顯著效益的影響。本研究以消費者的交易資料為探勘資料來源，每一筆交易資料包含消費者曾經購買的產品項目，以k位消費者為探勘目標，k≥1，以消費者為中心修改PAM（Partitioning Around Medoids）演算法，設計一個分群化交易資料成k個群組的方法，且分群化後之k個群組的產品相似度總和為最大，分別從各群組中找出目標消費者與產品項目之間關聯性，做為判斷k位消費者適性產品項目的依據。本研究分群化過程中除了保留PAM演算法的精神，取代原先中心點的交易資料也能具備各目標消費者的群組獨特性，且忽略與k位消費者無關聯的交易資料，可提升後續分群化計算效率。文中根據提出的方法，設計與建置一個發掘消費者適性產品項目探勘系統，本研究探勘結果對企業規劃消費者適性產品項目推薦，可以提供相當有用的參考資訊。

關鍵字

資料探勘 ； 分群化 ； PAM ； 適性產品

並列摘要

Enterprises can easily record and store consumer transaction records using information technology. If we can analyze consumer interest trends from quickly accumulating large amounts of data, it will has significant benefits to assist enterprises to provide consumers' adaptive products marketing strategies and enhance market niche. This paper uses consumers' transaction data as the source data of mining, and each transaction data contains the product items that a consumer has purchased. Let k consumers as the target of mining, k≥1. Considering the consumers as the center, we modify the PAM (Partitioning Around Medoids) algorithm to present a clustering method to cluster transaction data to k groups with the maximum similarity of products. The association between other products and the target consumers are found from each group, and as the basis for judging k individual consumers' adaptive products. In addition to keep the spirit of the PAM algorithm in the process of clustering, replacing the transaction data of the original center points also have the groups uniqueness with the target consumers. Ignoring transaction data unrelated to k consumers can improve the efficiency of subsequent clustering computations. A mining system of finding consumers' adaptive products is designed and built according to the proposed methods. The results of mining can provide very useful information to plan consumers' adaptive products marketing for enterprises.

並列關鍵字

Data Mining ； Clustering ； PAM ； Adaptive Product