UE Network Traffic Time-Series (Applications, Throughput, Latency, CQI) in LTE/5G Networks

Citation Author(s):
University of Thessaly, Sorbonne Université
University of Thessaly
University of Thessaly, CERTH Greece
University of Thessaly, CERTH Greece
Submitted by:
Theodoros Tsourdinis
Last updated:
Fri, 12/09/2022 - 09:43
Data Format:
1 rating - Please login to submit your rating.


This dataset includes real-world time-series statistics from network traffic on real commercial LTE networks in Greece. The purpose of this dataset is to capture the QoS/QoE of three COTS UEs interacting with three edge applications. Specifically, the following features are included:  Throughput and Jitter for each UE-Application and Channel Quality Indicator (CQI) for each UE. The interactions were generated from a realistic network behavior in an office by developing multiple network traffic scenarios. These scenarios are based on real network patterns observed at a specific time interval during the day (early morning from 10:00 AM to 11:00 AM) on users in our office facilities in Volos, Greece. The mobility of users is considered as well, since we developed attenuation scenarios from real commercial networks in Volos, Greece. These attenuation scenarios emulate cars traveling a specific city route with speeds that vary from 40 to 60 km/h. These car scenarios were used to collect 182500 CQI data from 73 cars capturing a large spectrum of the route's traffic. To attenuate the signal in order to emulate the realistic mobility scenarios, we utilized Programmable Attenuators that were connected directly to the RAN. The CQI dataset is publicly available here. Traffic monitoring, as well as traffic analysis and CQI, is captured/calculated in almost real-time from our custom Data Analytics Function (NWDAF) named Core & Ran Analytics Function (CRAF). CRAF utilizes the PyShark to live capture the traffic and FlexRAN controller to obtain RAN Statistics such as the CQI. Then it stores all the data in a MySQL database. These data were exported as a .csv file for easy data analysis and preprocessing.


Units of measurement of features:

  • Throughput: bytes
  • Jitter: seconds 

To utilize this dataset efficiently, we propose using the respective GitHub repository which contains:

  • The architecture of our framework.
  • The traffic/mobility scenarios.
  • Notebooks that include the utilized machine learning pipeline to develop various AI/ML Models that use this dataset.