Peer comparison is one of the most desirable features for network customers. One of the most frequently asked questions is, "How does my company's network performance compare with that of my peers?" To provide effective peer comparison results there are two fundamental questions that must be resolved – the first question concerns finding the most similar peers and the second question addresses understanding why the peers are similar. To address these types of challenges, techniques are presented herein that leverage machine learning (ML) models to resolve the two fundamental questions that were described above. Aspects of the presented techniques encompass an end-to-end system, which for convenience may be referred to herein as "DeepSense," which resolves the entire lifecycle mystery of peer comparison. Additionally, aspects of the presented techniques employ a singular value decomposition (SVD) algorithm to define similarity among customers in a way that is able to overcome the limitations that are caused by latent information. Further, aspects of the presented techniques leverage non-negative matrix factorization (NMF) to capture the dominant features which can influence the similarity among peers. Still further, aspects of the presented techniques support a user-friendly customer interface in real working production systems.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Zhang, Xinjun; Jung, Doosan; Shao, Qihong; Broich, Kevin; Liu, Yue; and Singh, Gurvinder, "DETERMINING LIKE PEERS AND DOMINANT FEATURES USING MACHINE LEARNING", Technical Disclosure Commons, (August 24, 2021)