Evaluating the Suitability of LSTM Models for Edge Computing

This study evaluates the suitability of using a long short-term memory (LSTM) neural network architecture to detect anomalies in a resource-constrained environment. It compares different hyperparameter configurations of LSTM models and evaluates their performance on a resource-constrained edge device. This study uses an edge testbed to evaluate the models and uses hybrid resources to train and update the LSTM models for faster development. The UNSW-NB15 dataset was used for model training and evaluation. The results demonstrate the effectiveness of the LSTM model in detecting anomalies in a resource-constrained environment. Additionally, this approach highlights the scalability and capability of LSTM at the edge and the use of a connected cloud for further analysis.