Predicting Service Faults On Telecommunications Network Using Machine Learning

Abstract

Fault severity prediction is a crucial business statistic for telecom providers to create customer advocacy. In this project, we use data mining techniques to predict fault severity at a location. We evaluate various machine learning models based on various classification metrics.

You can find my Proect Documentation here

Objective

To predict Fault Severity at particular time and location based on log data available Fault severity has 3 categories: 0,1,2 (0 meaning no faults, 1 meaning only a few faults, and 2 meaning many faults)

Overview of Simple End to End Architecture which I followed

Deployed Streamlit web app on Heroku Cloud created so that users could explore my LGBM Classification model

The app helps predict the Telstra network’s Fault Severity at a time at a particular location based on the log data available.

Steamlit App Deployed on Heroku