A FCM clustering algorithm based on Semi-supervised and Point Density Weighted

The effect of FCM depends on the samples´ distribution. The optimum clustering result might be not valid for the data sets having mass shape and large discrepancy of every class specimen number. Therefore, a Semi-supervised and Point Density Weighted Fuzzy C-means clustering (SSWFCM) is proposed. This algorithm using distance-based semi-supervised learning studies the training data set and gets coefficient matrix of each category, and then using the distance formula with a coefficient and point density weighted clusters the test data sets. The experiment proves that SSWFCM is superior to FCM in the clustering accuracy and validity. Moreover, the introduction of point density weight making SSWFCM can handle data sets with different distributions.