A study of K-Means algorithm for clustering problems with fuzzy linguistic attributes

• Main Authors: Lin Chieh-Ming, 林杰民
• Other Authors: Lee Hong-Tau
• Format: Others
• Language: zh-TW
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/09624139013587320676

碩士 === 國立勤益技術學院 === 工業工程與管理系 === 93 === The generalized k-means method is based on the measurements through a predetermined penalty function that measures the discrepancy of assigning data of a dataset to k clusters with k points as their centres. The k central points and the assignment are modified based on the intention of minimizing the measurements. This method has been applied to data analysis and the decision-making problems. Since some decision-making problems have to deal with the data that collected with imprecision attributes, we extend its application capability in this study. The linguistic variable of fuzzy set theory is considered. The k-means method is then modified so that it can handle such kind of data. For this purpose, we first have to make a deep research about the k-means method to realize the issues encountered by the method and their possible solutions. Second, we study the properties and operations for linguistic variable of fuzzy sets theory. Finally, the k-means method is modified to handle such kind of data and the theoretical issues of complexity and convergence properties are studied also. Our preliminaryresult indicates the modification of k-means algorithm is feasible. The additional research about the theoretical issues and application of real data is necessary for verifying the proposed approach.