Big Data Enabled Anomaly User Detection in Mobile Wireless Networks

With the rapid development of the mobile Internet and the Internet of Things (IoT), mobile data traffic has been exploded. Wireless communication networks have entered the era of big data. Anomalous user can be studied with their negative experience by analyzing users' activities in wireless networks. In this paper, we propose a novel mobile big data (MBD) architecture consisting of four layers, including storage layer, fusion layer, analysis layer and the application layer. Based on the MDB architecture, we present a data-driven user experience prediction as a case study of applying the proposed MBD architecture in wireless network. By leveraging machine learning algorithms, the proposed user experience prediction can pre-evaluate user experience through network performance and user behavior features in a data-driven fashion. First, we perform a preliminary analysis on consumer complaints records obtained from the network monitoring system of a major mobile network operator (MNO) in China. Second, up-sampling and down-sampling are combined to combat the severe imbalanced negative and positive samples. The results show that proposed automated machine learning algorithm improves the prediction accuracy compared with two commonly baselines adopted by the MNO: empirical criterion and rule-based expert system.