Committed to connecting the world

AI for Good Global Summit


​​​​​Radio Link Failure Prediction​ Challenge

banner_ZTE_Cisco_2.png Turkcell is submitting a problem statement to the ITU Artificial Intelligence/Machine Learning in 5G Challenge, a competition which is scheduled to run from now until December, 2020. P​articipation in the Challenge is free of charge and open to all interested parties from countries that are members of ITU.

Detailed information about it can be found on the Challenge website, which includes the document “ITU AI/ML 5G Challenge: Participation Guidelines”.

Radio Link Failure Prediction Challenge is organized as a part of ITU AI/ML in 5G Challenge.

Challenge: Using weather info for radio link failure (RLF) prediction


Cloud, rain, snow, and other weather-related phenomena affects the performance of radio links. This is especially applicable to backhaul links operating at GHz frequencies. A generic regional weather forecast data is available which lists expected conditions and coarse temperatures along with actual –precise– realizations.

Adding to the complexity are the spatial nature of the data (regions of weather data and RLF needs to be aligned) as well as the time sync needed to correlate various occurrences. Over a period of time, we have compiled and anonymised region-wise data which corresponds to weather forecasts, and RLFs derived from our networks.


Given the region-wise, historical data sets derived from our networks, with weather forecast as well as radio link (RL) performance (for a given frequency band), predict the RLFs. 

​Data source

Training data will include pre-processed and anonymised RL KPIs from our networks and time-aligned weather data. RL KPI data includes date/time, coordinates, frequency band, link length, error and failure statistics, availability ratio, stability score, capacity, modulation (128QAM, 256QAM, 512QAM, etc.).

Weather forecast data includes coordinates, temperatures (min, max), humidity (min, max), wind speed and direction while the hourly weather realizations data includes precipitation and overcast ratio in addition to them.

Weather forecast data is provided twice per day (one for morning hours and one for evenings hours) for the following 5 days where the realizations are recorded hourly.


  1. Please check the slides here on what to submit

Evaluation criteria:​

  • Participants must use the provided data set to train a machine learning algorithm.
  • The output of the ML algorithm should be able to predict the performance obtained in a new network deployment. 
  • The choice of the ML approach is decided by each participant.
  • A test data set will be provided to evaluate the performance of the proposed algorithms.
  • The evaluation of the proposed algorithms will be based on the f1 scores of the radio link failure predictions​.
  • The winners will be given prizes (and may be invited to publish the results in an academic publication or present in a conference, etc).​


  1. Registration [closed]31 July 2020 21 August 2020 
  2. Global Round duration: July - November 2020
  3. Training & Validation data sets: Available July 15
  4. Test data set: will be released later.​ (participants to submit models which will be  evaluated against the test dataset)
  5. Deadline to submit a solution: 15 October 2020 30 October 2020​ (Extended)​
  6. Announcement of the winners: November, 2020​


Data:Dataset can be downloaded here (data set has been updated)

How to participate?

  1. If you don't have an ITU account, please follow the guidance to create one for challenge registration.
  2. Register on ITU AI/ML in 5G challenge website with your ITU account.
  3. Fill out the ITU AI/ML in 5G Challenge Participants Survey​​ to select problem statement ITU-ML5G-PS-036. You can enroll as a team with 1-4 members.
  4. Begin to work on this problem and submit your results. 
  5. This problem statement is open to ITU members and any individual from an ITU Member State


ITU contact: ai5gchallenge[at]

​Radio Link Failure Prediction​ Challenge contact:

Aydin Çetin, aydin.cetin[at]

Serkan Karadag, serkan.karadag[at]

Sinem Çakmak Gürsel, sinem.cakmak[at]

Salih Ergüt, salih.ergut[at]

Ismail Hakki Ozcelik ismail.ozcelik[at]  

Bahar Ozturk bahar.turkmen[at]​

Ilker Bilge​  ilker.bilge[at]

Semih Aktas semih.aktas[at]

​You can also visit our Slack Channel​ to find more guidance.