Similarity measurement for data with high-dimensional and mixed feature values through fuzzy clustering
Liu Haitao, Wei Ru-xiang, Jiang Guo-ping · 2012
For data with high-dimensional and mixed feature values, traditional similarity measurement becomes no longer applicable. In this paper, a new similarity measurement is proposed by designing a high dimension FCM clustering algorithm. Firstly, an initialization of ordinal-numerical mappings is given; secondly, new ordinal-numerical mappings are learned from the iterative high dimension FCM clustering algorithm and the clustering effect becomes optimized at the same time; finally, a new similarity measurement for data with high-dimensional and mixed feature values is proposed with the fuzzy partition matrix. Experimental results show that the similarity measurement improves the precision of estimation.