Abstract: Because traditional clustering algorithms are difficult to deal with large-scale data and sensitive to noise data, based on the Fuzzy C-ordered-means clustering (FCOM) algorithm, we propose a single-pass fuzzy C-ordered clustering algorithm, named SPFCOM, and an online fuzzy C-ordered clustering algorithm, named OFCOM, by combining single-pass and online incremental architectures respectively. These two algorithms weight the FCOM algorithm, and incrementally process the large-scale data chunk by chunk. Experimental results show that, compared with other similar prominent algorithms, the SPFCOM and OFCOM algorithms can achieve higher accuracy and better robustness.

Key words: fuzzy clustering, incremental clustering, robustness