Identifying Cell Sector Clusters Using Massive Mobile Usage Records

Abstract

Optimizing capital expenditure (CapEx) has been an increasingly important objective in telco operators’ cell planning process. Traditionally, neighbor cell relation is operationally managed and independent from capacity planning. In this paper, we present SCUT, an algorithm that uses massive mobile usage records to detect clusters of possible capacity-sharing sectors, such that capacity planning can be optimized based on coverage. SCUT analyzes shared usage to build a graph-based model of an operator’s network and identifies its disjoint dense components as best-fit abstractions of clusters. Through analysis and benchmarking on real data, we demonstrate its scalability and potential to improve industry-standard site-based planning. SCUT has been deployed for a telco operator in Southeast Asia.