Deep Learning-Based Big Data-Assisted Anomaly Detection in Cellular Networks

5G is envisioned to have an artificial intelligence (AI)-empowerment to efficiently plan, manage and optimize the extremely complex network by leveraging colossal amount of data (big data) generated at different levels of the network architecture. Cell outages and congestion pose serious threat to the network management. Sleeping cell is a special case of cell outage in which the cell provides inferior services to its users. This peculiar behavior of the cell is particularly challenging to detect as it disguises itself from the network monitoring entity. Inadequate accuracy and high false alarms are two major constraints of state-of-the-art approaches for the anomaly-sleeping cell and surge in user traffic activity that may lead to congestion-detection in cellular networks. This implies squandering of scarce resources which ultimately results in increased operational expenditure (OPEX) while disrupting network's quality of service (QoS) and user's quality of experience (QoE). Inspired from the prominent success of deep learning (DL) technology in machine learning domain, this is the first study that applies DL for the detection of abovementioned anomalies. We utilized, and did a comprehensive study of, L-layer deep feedforward neural network fueled by real call detail record (CDR) dataset (big data) and achieved 94.6% accuracy with 1.7% false positive rate (FPR), that are remarkable improvements and overcome the limitations of the previous studies. The preliminary results elucidate the feasibility and preeminence of our proposed anomaly detection framework.