關於電信門號資費的相關研究文獻，過去研究大都以使用者的不同需求亦或針對電信系統商的門號專案規劃提出建議或分析模型。相關文獻皆未以電信零售商的角度進行探討。因此，本研究將以電信零售商所面臨的問題為研究議題提出一混合K-means與C5.0之分類反應模型。先透過K-means分群方法將資料先做分群分析。接著透過C5.0分類方法針對分群結果建立分類法則。藉由分群法則將可以針對分群結果做有效的表達說明。由實驗結果可發現本研究所使用之模型可有效的分析商品與佣金組合，經由10-fold之分類準確率為97.67%。當未來有新的資費專案產生時亦能夠透過此模型進行判斷所屬類別。透過本研究所提出方法可用以協助零售商找尋互補性的門號資費商品。當商品缺貨或是新品上市時就可以快速了解該項商品可搭配的資費與專案。亦可以快速了解高毛利的資費商品，將可以針對此類型商品做為主力推銷的商品。;The literature on telecom tariff mainly focused on the meeting users need with appropriate telecom tariff. However, no studies had analyze product benefit from the perspective of retail channels. In this study, the hybrid K-means and C5.0 classification model have been used to investigate the factors attesting tariff in mobile device retail channels. The K-means method was used for clustering products according to costs and benerfit, and then the C5.0 classification method was applied to analyze the characterstics of products in each cluster. The classifation results were presented as a set of rules. The experimental results indicated that the proposed approach showed excellent performance with accuracy of 97.76% in 10-fold cross validation. With the rules, telecom retailers could recommended right substitute products for out of stock items and classifying new arrivals to correct product clusters.
