Poor Data Throughput Root Cause Analysis in Mobile Networks using Deep Neural Network
Maluambanzila Minerve Mampaka, Mbuyu Sumbwanyambe · 2019
As the mobile data usage is increasing with the complexity of new services, ensuring proper quality of experience (QoE) becomes challenging. While traditional machine learning techniques have been intensively applied in mobile networks, new training models are also mushrooming and are to some extent explored to handle large amount of data and complexity of new services in mobile data networks. Lately, deep learning algorithms have gained popularity since they scale efficiently with large data input and make use of the computing power to train the models. This paper proposes a deep learning model based on a deep neural network (DNN) architecture to train and evaluate the root cause analysis (RCA) for poor throughput in mobile networks. The proposed approach considered both the radio and the core network performance indicators of a mobile data network as inputs to ensure end-to-end correlation. Furthermore, the local interpretable model-agnostic explanations (LIME) method was used to provide features importance both for the global system and for individual subscribers. The prediction accuracy on the unseen data was 98.9% with an area under the receiver operating characteristic (ROC) curve of 99.87% and an F1_score of 99.23%.