An Urban Multi-Operator QoE-Aware Dataset for Cellular Networks in Dense Environments
Description
This dataset contains 30,925 labelled and cleaned records collected from a dense 2 km² urban area surrounding Sunway University, Selangor, Malaysia. Using the GNetTrack Pro mobile application and Samsung S21 Ultra devices, the data spans three anonymized commercial mobile network operators and includes both 4G and 5G technologies. The dataset captures radio signal quality metrics (RSRP, RSRQ, SNR, etc.), geospatial information, mobility patterns (walking vs. driving), and application-specific traffic scenarios (HTTP, FTP, 1080p Video Streaming). A total of 132 physical cell sites were validated via OpenCellID and field inspections. The dataset is released in CSV format and includes Python scripts for data preprocessing and basic visualization. This makes it a valuable resource for machine learning tasks like signal metric regression, handover optimization, and QoE modeling in heterogeneous and simulation of high-density urban networks. Key features: Real-world 5G/4G measurements Multi-operator and multi-mobility modes Traffic-aware profiling Empirical validation of base station locations Ready for ML/DL use cases For more information look out for our article about the dataset on data in brief journal.
Files
Steps to reproduce
Step 1: Data Collection Tool Used: GNetTrack Pro (Android app by Gyokov Solutions) Device: Samsung S21 Ultra Location: Outdoor areas within 2 km radius of Sunway University Operators: 3 major Malaysian operators (anonymized as A, B, C) Mobility Modes: Walking (pedestrian/canopy) and Driving (shuttle/BRT) Traffic Contexts: HTTP browsing, FTP uploads/downloads, 1080p video streaming Upload: Logs were auto-uploaded to a cloud database Step 2: Data Preprocessing Download raw logs from the cloud. Merge and clean data using python script: Remove duplicates and null values Validate and cap signal metrics using 3GPP TS 36.214 standards Categorize mobility (Walking, Driving) using speed thresholds Map and anonymize operator names Engineer features: Mobility from speed Node_Latitude / Node_Longitude using signal centroid mapping ElapsedTime for session-based modeling SessionID for sequence tracking Resulting output: raw_dataset.csv Step 3: Data Analysis & Visualization Use Colab notebook to: Analyze network technology distribution Visualize signal strength distributions Plot traffic types and altitude profiles Estimate node locations using centroid-based techniques Step 4: Ready for ML/DL Cleaned data supports regression and classification tasks: Handover management Signal quality forecasting Multi-operator performance analysis Context-aware optimization for 5G networks