Machine learning based models for fault detection in automatic meter reading systems

Recently, research has focused on the area of fault detection in Automatic Meter Reading (AMR) systems. The manufacturers and users of AMR systems are now keen to include diagnostic features in the systems to improve salability and reliability. However, traditional manual fault detection methods are time-consuming and inaccurate. automatic fast fault detection methods are urgently needed. In this paper, we propose several machine learning based fault detection models to meet this requirement. Furthermore, we use novel boosting strategy to fuse multiple models to leverage multi-aspect information in AMR systems. The experimental results on simulated data show that the proposed models are accurate and robust, and fusion strategy indeed improve the performance on fault detection.