LEAF: Navigating Concept Drift in Cellular Networks

Operational networks commonly rely on machine learning models for many tasks, including detecting anomalies, inferring application performance, and forecasting demand. Yet, unfortunately, model accuracy can degrade due to concept drift, whereby the relationship between the features and the target prediction changes due to reasons ranging from software upgrades to seasonality to changes in user behavior. Mitigating concept drift is thus an essential part of operationalizing machine learning models, and yet despite its importance, concept drift has not been extensively explored in the context of networking -- or regression models in general. Thus, it is not well-understood how to detect or mitigate it for many common network management tasks that currently rely on machine learning models. As we show, concept drift cannot always be mitigated by periodic retraining models using newly available data, and doing so can even degrade model accuracy. In this paper, we characterize concept drift in a large cellular network for a metropolitan area in the United States. We find that concept drift occurs across key performance indicators (KPIs), regardless of model, training set size, and time interval -- thus necessitating practical approaches to detect, explain, and mitigate it. To do so, we develop Local Error Approximation of Features (LEAF). LEAF detects drift; explains features and time intervals that most contribute to drift; and mitigates drift using resampling, augmentation, or ensembling. We evaluate LEAF against industry-standard mitigations (i.e., periodic retraining) with more than three years of cellular data from Verizon. LEAF consistently outperforms periodic retraining on a variety of KPIs and models, while reducing costly retrains by an order of magnitude. Due to its effectiveness, a major cellular carrier is now integrating LEAF into its forecasting and provisioning processes.