Radio-Link-Failure-Prediction

Research paper: https://www.itu.int/pub/S-JNL-VOL3.ISSUE2-2022-A13

An ML-based RLF Prediction Algorithm to predict failures in a radio-link wireless communication channel

Stable and high-quality internet connectivity is mandatory to 5G mobile networks, but once something unexpected happens, the influence of the defect is quite severe. This is due to various weather-related phenomena that affect the performance of radio links like clouds, rain, snow etc. Thus correct radio link failure prediction is necessary to avoid such radio link failures.

The objective of this problem statement is to predict the occurrence of radio link failures in:

1. in the next day
2. in the following 5 days
   

A Machine Learning and Time Series Forecasting model is built for the same.

The Repository contains the following files:

1. Training_Validation_Notebook.ipynb: Python notebook with code for training and validation
   
2. final_feature_engineering.ipynb: Python notebook with code to predict the redundant columns that can be eliminated to improve performance of failure prediction
   
3. Training_preprocess.ipynb: Python notebook to preprocess training data
   
4. validation_preprocess.ipynb: Python notebook to preprocess validation data
   
5. 20210125_predicts.tsv: TSV file containing validation output
6. 20210426_predicts.tsv: TSV file containing validation output 1
7. 20210525_predicts.tsv: TSV file containing validation output 2
8. 20210614_predicts.tsv: TSV file containing validation output 3
9. 20210817_predicts.tsv: TSV file containing validation output 4

The problem statement is a part of the ITU AI for Good Machine Learning in 5G Challenge. The dataset was provided by Turkcell and can be accessed using the following link: https://github.com/Turkcell/ITU-AIMLin5GChallenge-2021

Zip file of the dataset contains the following tab separated files (tsv):
• distances.tsv: pair-wise distances
• met-forecast.tsv: meteorology 5-day forecasts
• met-real.tsv: meteorology historic realizations
• met-stations.tsv: meteorology station information
• rl-kpis.tsv: radio link KPIs and configuration parameters
• rl-sites.tsv: radio link site information

Citation of the paper

@article{Nethraa_Sivakumar_2022,
doi = {10.52953/lzlj8762},
url = {https://doi.org/10.52953%2Flzlj8762},
year = 2022,
month = {jul},
publisher = {International Telecommunication Union},
volume = {3},
number = {2},
pages = {142--156},
author = {Nethraa Sivakumar and Pooja Srinivasan and Nikhil Viswanath and Venkateswaran N},
title = {Decision tree-based radio link failure prediction for 5G communication reliability},
journal = {{ITU} Journal on Future and Evolving Technologies}
}