Friday, May 15, 2020
Mobile Coverage Maps Prediction
Manos Alimpertis, UCI
11:00am, Zoom (link to be sent via email) [Host: Scott]
Mobile coverage (signal strength) maps are of great importance to cellular operators, however they are expensive to obtain, inaccurate in some locations, imperfectly reflective of call quality outcomes and potentially constructed from biased samples. In this talk, we develop a principled machine learning prediction framework to address these challenges. First, we develop a prediction framework based on random forests (RFs) to predict signal strength maps from limited measurements. The proposed RFs-based predictor utilizes a rich set of features including but not limited to location, time, cell ID and device hardware, which are considered jointly for the first time. Our method improves the tradeoff between prediction error and the number of training points compared to prior work. Second, we extend the framework to (i) predict quality outcomes beyond signal strength (e.g., coverage indicator or call drop probability) and signal maps optimized beyond MSE; and (ii) deal with sampling bias using ideas from importance sampling. Third, we apply, for the first time, data Shapley valuation to mobile coverage map prediction, and we demonstrate how it can be used for data minimization in this context. Throughout this thesis, we leverage two types of real-world LTE datasets to evaluate our methods and gain valuable insights: (i) a small but dense Campus dataset, collected on UC Irvine campus by ourselves and (ii) several large but sparser NYC and LA datasets, with approx. 11M, provided by a mobile data analytics company.
Emmanouil (Manos) Alimpertis received his Diploma (5 year degree) and M.Sc. in Electronic and Computer Engineering at the Technical University of Crete in Greece, in 2012 and 2014 respectively. He joined the Networked Systems PhD Program in University of California, Irvine in 2014 and has since been working under the supervision of Professor Athina Markopoulou. He has held research internship positions at AT&T Research labs (summer 2015 and summer 2017) and at M2Catalyst mobile analytics (summer 2016) as well as a software engineer intern at Apple (summer 2019). His research interests span the area of crowdsourcing, cellular mobile networks, machine learning for networking applications and location estimation algorithms. He is expected to graduate at the end of the spring quarter and is grateful for the opportunity to practice his defense talk!