FCM Clustering And Research Of Its Increment Algorithm

As an important branch of the data mining technology, clustering analysis is a kind of very important method. It studies data between physical or logical relationship, the dataset is divided into certain classes of similar nature data points by special rules.Fuzzy C-means algorithm (FCM algorithm) is one of widely application of the algorithm. It does this by repeatedly revised membership matrix and the cluster center, so as to achieve the purpose of classification of the samples.Since FCM algorithm is influenced by the initial cluster centers, there is a large number of improved algorithm at present. This paper discusses the fuzzy c-means clustering and the incremental clustering algorithm.The main work is as follows:1. Analyzes the advantages and disadvantages of FCM clustering algorithm, the Laplace index is introduced into the objective function, and the structural information between the clustering object is automatically converted to the weight of the object to improve the accuracy and speed of the algorithm. The improved algorithm combined with the incremental clustering algorithm, this method can avoid a lot of repetition computation, and unaffected by isolated point.2. Analyzes the existing semi-supervised fuzzy c-means clustering algorithm, proposed a new improved semi-supervised fuzzy c-means clustering algorithm. The algorithm combines FCM algorithm based on ant colony and semi-supervised algorithm, improve the performance of the algorithm by changing the objective function.Accounted for marked samples of all samples, the greater proportion, the higher performance.3. The improved algorithm achieve with the MATLAB programming by select some datasets. Compared with the results of other algorithms, and analysis the results of the experiment.