Multi-Operator Dataset for Throughput Prediction Across Diverse Mobility Modes

This dataset features real-world radio frequency (RF) measurements and network performance metrics collected within a dense urban and academic ecosystem in Sunway City, Malaysia. The data was engineered to support research in mobile network performance, user equipment (UE) behavior, and the predictive modeling of user-level throughput in 5G New Radio (NR) and LTE environments. The data collection architecture utilized commercial Samsung Galaxy S24 smartphones interfaced with the Keysight Nemo Handy drive test tool to capture high-fidelity engineering logs directly from the baseband processor. The dataset uniquely captures concurrent performance metrics across three top-tier mobile network operators. To facilitate research into realistic user experiences, data collection spanned three distinct micro-mobility environments reflective of modern smart-city infrastructure: Canopy Walks: Representing low-speed pedestrian mobility and elevated walkways. Shuttle Buses: Representing medium-speed street-level transit. Elevated BRT (Bus Rapid Transit): Representing dedicated, higher-frequency elevated transit corridors. Records include highly granular, sub-second tracking of spatial parameters (Latitude, Longitude), radio link quality (RSRP, Band, Physical Cell Identity), and network layer attributes (Beam Index, Cell Type). The primary target variable for machine learning frameworks and time-series forecasting is Throughput_Mbps, captured during active network sessions.本数据集收录了马来西亚双威城高密度城区与学术园区生态内采集的真实世界射频（RF）测量数据与网络性能指标。该数据集专为支撑5G新空口（NR）与长期演进（LTE）环境下的移动网络性能、用户设备（UE）行为及用户级吞吐量预测建模相关研究而构建。 本次数据采集采用商用三星Galaxy S24智能手机，搭配是德科技（Keysight）Nemo Handy路测工具，可直接从基带处理器捕获高保真工程日志。本数据集的独特优势在于，可同时采集三家顶级移动网络运营商的性能指标。为助力真实用户体验相关研究，数据采集覆盖了三类贴合现代智慧城市基础设施的典型微移动场景： - 林荫步道（Canopy Walks）：对应低速步行移动与高架步道场景。 - 穿梭巴士（Shuttle Buses）：对应中速地面公交场景。 - 高架快速公交（BRT，Bus Rapid Transit）：对应专用高频次高架公交专用道场景。 采集记录包含高粒度的亚秒级跟踪数据，涵盖空间参数（纬度（Latitude）、经度（Longitude））、无线链路质量指标（参考信号接收功率（RSRP）、频段、物理小区标识（Physical Cell Identity））以及网络层属性（波束索引（Beam Index）、小区类型（Cell Type））。面向机器学习框架与时间序列预测任务的核心目标变量为活跃网络会话期间采集的吞吐量（Throughput_Mbps）。