Predicting the fault severity of the network at a time at a particular location based on the features extracted from log data, joined from 5 different CSV files. Main dataset had 8K rows train, 12K rows test, and features were combined from the different files.It is a multinomial classification model, which predicts no disruption, momentary glitch or total interruption of connectivity. This model will enable the company to predict in advance the scope and timing of service disruptions.
Metrics: Multi-class Log-loss: 0.779