臺灣博碩士論文加值系統

本論文是探討模糊資料之聚類方法，內容重點可分為模型架構與實際應用兩部份。在模型架構部份，我們提供了一個在h-切割下之區間來進行聚類的方法，使用此方法來對模糊資料聚類，可得知這群資料最均質之聚類情形 (h=1時)，也可知道在其他h值下的各種聚類情形，以供比較參考。同時，當我們已知一真正類別的資料時可將其模糊化後，若其在某個h值之上的聚類結果與未模糊化時之分類結果相同，則往後只要所收集到的資料是介於此h-切割之區間內，就可正確的將之歸類。在實際應用方面，我們以中華電信公司所要提供給大眾的十種網路服務項目，針對其技術難易程度（模糊資料），將其歸為三類（期），而得知在各期所要發展的服務項目。

In this thesis , we proposed a method to solve a general clustering problem of which the data is fuzzy. There are two major parts in this thesis: one is model-development and the other is a practical application. Regarding the model-development, we extended the Bi-Objective Fuzzy C-means method to the one that can classify fuzzy data by the interval of any h-cut. When we use the proposed method, we not only have the most homogeneous classification(when h=l), but also have different clusterings from different h values. As for the practical applications , we have to classify ten potential services provided by Broad ban information network into three clusters, so that these services in the same cluster can be developed simultaneously. Besides, if we have known the actual cluster in which all the crisp data could belong to, then we can fuzzify these crisp data. If the results of classifying these fuzzified data at certain h-level is the same as that of the original crisp data, then once a collected datum falls in this h-level, we can identify its belonged cluster.