This project aims to resolve the problem of data loss for energy status monitoring system on server side due to limitations of operations, or uncontrollable user existence. A solution is proposed in the paper to provide a high accuracy of estimation on energy status in large-scale beacon networks which supports a small amount of existing records.


  • 1. Proposed a novel SVR-based framework that can cope with incomplete training data and adapt to varying envi-ronment, configuration, and hardware types
  • 2. Proposed a novel recurrent training method that can capture both local and global features for high accuracy
  • 3. Collected real-life dataset and used it to prove the effectiveness of the proposed framework


The dataset for the simulation are available in our Github repo.