A Study on Density-Based Data Management in Edge Computing and Its Application to Anomaly Detection

Abstract

In recent years, wide spread of IoT has made it possible to acquire enormous amounts of sensor information and artificial intelligence technologies has made dramatic progress by utilizing this information. As explosive increase in such data volume, it becomes difficult to collect and process all data in one place. Therefore, storing and processing data on edge side is becoming important. However, edge devices usually have only limited computation and memory resources and hence, it is not practical to save all the acquired data. There is a great demand to select the data to be stored effectively at the edge. In this paper, we propose an efficient density-based data management technique. We also propose an on-line anomaly detection system that applies the proposed data management technique with sequential learning and periodic retraining. Throughout experiments, we found that our system achieved higher accuracy than conventional data management techniques.