UCK-means: A customized K-means for clustering uncertain measurement data

Due to some reasons such as transmitting error or outdated or imprecise measurement, data uncertainty is an inherent property in wireless sensor networks or in LXI test framework. When we apply data mining techniques to these uncertain data, we must consider the uncertainty to get better data mining results. At present, most of uncertain data clustering methods assume the probability density functions or probability distribution function of whole data is available. However, in many real applications, this piece of information is rarely available. Only limited uncertain information may be available, such as the standard deviation. In this paper, we adopt a more realistic assumption that the standard deviation of individual measurement data is available, and propose a new uncertain distance computing method between multi-dimensional uncertain data. In addition, we propose an uncertain customized data clustering algorithm based on the classical K-means to process the multi-dimensional uncertain data. Experiment results show that the uncertain clustering algorithm can produce better results with lower complexity.