Mobile Networks Classification Based on Time-Series Clustering

With the increasing complexity of network architecture, the classification of mobile network cells become more important in network operation and maintenance. However, the previous classification method based on manual annotation of scene labeling is inefficient and biased. In this paper, we focus on proposing a data-driven classification method to eliminate the drawbacks of manual annotation. The proposed method extracts the patterns of mobile network on temporal shape and statistical features, and then calculates the fused distance matrix from these two feature sets, K-medoids is leveraged to get the classification labels. We design a series of experiments and analyses to demonstrate the validity of the proposed method, which is based on hourly real data sampled from 9454 mobile cells. The experiments demonstrate that the proposed method achieves good performance on the cell classification of two O&M (Operations and Maintenance) scenarios, and significantly improves the work efficiency.