A New Belief-Based Incomplete Pattern Unsupervised Classification Method

The clustering of incomplete patterns is a very challenging task because the estimations may negatively affect distribution real centers and thus cause uncertainty imprecision in results. To address this problem, new belief-based pattern unsupervised classification method (BPC) proposed paper. First, complete are grouped into few clusters by classical soft like fuzzy $c$ -means to obtain corresponding reliable thereby partitioned unreliable ones an optimization method. Second, basic classifier trained employed classifies edited neighbors. In way, most can be submitted specific clusters. Finally, some ambiguous will carefully repartitioned again distance-based rule depending on obtained belief functions theory. By doing this, that difficult classify between different reasonably meta-cluster which characterize due missing values. simulation results show BPC has potential deal with datasets.