A Hybrid Model with CNN-LSTM for Link Quality Prediction

Link quality prediction significantly improves network performance. Effective link quality prediction can ensure data transmission, and high-quality link transmission data can improve throughput, thereby extending the lifespan of the network. In order to better extract the temporal features of link quality parameters, this paper proposes a deep learning model that combines one dimensional convolutional neural network (IDCNN) and long short-term memory (LSTM). Firstly, local features are extracted from the original sequence using one dimensional convolution network. Secondly, temporal features are extracted using the sequence modeling advantages of LSTM. Finally, attention mechanism is introduced to improve the prediction performance of the model further. Experiments on real and public datasets have shown that the proposed CNN-LSTM has better predictive performance compared to shallow machine learning models and traditional sequence models.