A Clustering Approach Using Enhanced K-Means in 5G Networks

Abstract

There have been amount of messages for multiple users that exist in 5G communication networks. In order to provide high-quality services for these users at the same time, the messages for different users need to be grouped by clustering. The clustering approach has been promoting the growth of smart businesses in transmitting applications. For a large-scale data, applying this mining algorithm is severely limited. In this paper, we focused on the K-Means clustering where the Frequent Pattern-growth (FP-growth) algorithm applied to each cluster. The mined frequent items of the short text always have the similar meaning. And the representative candidates based on frequent items can represent the whole cluster of short texts that they belong to. This method of an enhanced K-Means based on FP-Growth breakthrough the limit of assuming the number of clusters, k, known in advance. It can automatically find as many candidates of short sentence as possible, and runs much faster and consumes much less main memory than the general algorithm by the computing distances between text documents like Earth Mover’s Distance based on Word2Vector.