Incomplete data fuzzy C-means method based on spatial distance of sample
Jingyi Cao, Chongquan Zhong, Dan Li · 2019
In this paper, an incomplete data fuzzy C-means method based on spatial distance of sample(FCM-SDS) is proposed. The method uses the nearest neighbor rule to utilize the spatial distribution information of the sample to obtain the missing attribute information of the incomplete data to interpolate the incomplete data set. Combining the nearest neighbor samples, the spatial distance with the adjustment factor $\alpha$ is constructed to replace the original Euclidean distance, the nearest neighbor density is added to the cluster objective function in a weighted manner, and the clustering effect is improved. The method is validated on the Iris and Wine datasets of the standard database UCI. When the attribute missing ratio is increased from 0% to 20%, the method has the smallest misclassification compared with other methods.