Fault Prediction in Optical Transport Network using Machine Learning Algorithms - Case of Ethio Telecom
No Thumbnail Available
Date
2022-01
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
The availability of Optical Transport Network (OTN) is an important factor to provide dependable telecom services. This backbone infrastructure of ethio telecom suffers from frequent outages and fault alarm data are exploding in magnitude. Hence, in this research machine learning based fault interarrival time prediction models have been developed using Artificial Neural Network(ANN), Long Short-Term Memory(LSTM) and Gated Recurrent Unit(GRU) algorithms to select the best model that captures the overall fault interarrival pattern and help take preventive actions before its occurrence. Further, the models have been trained using 70/30 and 80/20 training schemes to study their stability and perform models comparison based on five months fault history data. The prediction performance of these models have been evaluated using Mean Squared error (MSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error(MAPE) and R-Square (R²).
The models prediction performances indicate that the ANN model has predicted the fault interarrival time with 16.08% MAPE while GRU shows 16.23% MAPE and LSTM model has 17.93% MAPE on average. The R² score for the ANN model falls between 0.2576 and 0.4923 followed by the GRU model that exhibits its score in between 0.2387 and 0.4912. For LSTM model its R² score lies between 0.1583 and 0.4641. Therefore, the ANN model is able to catch the general trend of the fault interarrival time followed by the GRU model. In addition, the average computational time taken per epoch by ANN model is 2.44 ms, GRU model consumes 10.22 ms and LSTM takes 13.94 ms. Hence, ANN model outperforms in terms of prediction performance and computational efficiency.
Description
Keywords
Fault Prediction, Optical Transport Network, Network Management System (NMS), Machine Learning, Artificial Neural Network (ANN), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU)