Unsupervised Deep Transfer Learning for Fault Diagnosis in Fog Radio Access Networks

Abstract

The rapid development of the Internet of Things with the requirements of ultrareliability and ultralow latency has imposed huge challenges on the radio access network operation and maintenance. Using artificial intelligence technologies can provide the accurate fault diagnosis rapidly and efficiently, but it is usually hampered by the lack of historical data as well as the certified fault labels. To deal with these challenges, in this article, an unsupervised deep transfer learning-based fault diagnosis method in fog radio access networks is proposed. Specifically, a transfer learning-based density-based spatial clustering of applications with noise method is first utilized to detect and label fault data in each interval by using the core-level information. Then, an unsupervised deep transfer learning method combining a convolutional neural network with a domain adversarial neural network is applied to classify the categories of unlabeled fault data by using cell-level information. The experimental results show that the proposed method can reduce the missed detection rate than the traditional method, and has better fault diagnosis accuracy than the reference methods.