A Measurement-Driven And Capacity-Aware Framework For 5G NR NSA Deployment

Abstract

Conventional LTE-to-5G NR Non-Standalone (NSA) planning remains predominantly coverage-oriented and often fails to capture user-experienced throughput degradation under realistic traffic conditions. This limitation is critical in NSA architectures, where LTE acts as the anchor layer for control and mobility while 5G NR provides additional capacity. This study proposes a measurement-driven, throughput-centric spatial framework to identify LTE Capacity Bottleneck Zones (CBZ) as a basis for more realistic LTE-to-5G NR NSA deployment planning. The main novelty is the integration of a KPI-weighted RF Index with Kernel Density Estimation (KDE), DBSCAN spatial clustering, and fuzzy spatial zoning to generate throughput-aware capacity maps rather than purely coverage-based assessments. Drive-test measurements were conducted in Lubuk Alung District, Indonesia, under live LTE network conditions, yielding 8,355 radio KPI samples (RSRP, SINR) and 25,613 HTTP downlink throughput samples with geolocation. Statistical analysis using Pearson/Spearman correlation, polynomial regression, and Random Forest regression reveals consistently weak relationships between RSRP/SINR and throughput, indicating that radio-layer indicators alone provide limited explanatory power for user-experienced performance. The proposed framework classifies the study area into three spatial zones: LTE Stability Zone (28.99%), LTE Degradation Zone (63.02%), and LTE Capacity Bottleneck Zone (7.99%), where CBZs are characterized by acceptable radio conditions but localized throughput degradation. These findings enable a shift from coverage-centric evaluation toward targeted, throughput-aware capacity optimization for LTE-to-5G NR NSA deployment planning