Hybrid Neural Network to Impute Missing Data for IoT Applications
Nabeel Al-Milli, Wesam A. Almobaideen · 2019 IEEE Jordan International Joint Conference on Electrical Engineering and Information Technology (JEEIT) · 2019
The new development of the Internet of Things (IoT) depends on reliable data delivery, where transferring data between devices should be accurate and fast to ensure high performance for IoT applications. IoT applications could suffer from low quality of data delivery due to several factors such as connection errors, sensor faults, or security attacks. Low quality of data delivery reduces the performance of IoT applications since if the collected data is incomplete it could eventually be useless. In this paper, we propose a hybrid neural network with genetic algorithm to impute the missing data for medical IoT applications. A deep learning neural network (Jordan network) is used as a model to predict the missing data, while the genetic algorithm is adopted to optimize the weights of the neural network. The obtained results show that the proposed algorithm is able to impute missing data with high classification value based on Area Under the Curve (AUC) and improve the final performance of IoT application by up tp 5%.