Research and Application Based on CNN-LSTM

In this paper, a deep learning network structure based on convolutional neural network and long short-term memory (LSTM) neural network is proposed. Combining 18 independent variables of known datasets, correlation analysis is performed, and after variable analysis, feature fusion method is used to extract shallow features and deep features and connect them using convolutional network, and the obtained vector information is input into LSTM units by convolutional fusion of features, and the results of each network are weighted and fused. The CNN-LSTM model established in this paper performs prediction and finally obtains a goodness-of-fit of 0.9381, which is better than the other models compared, and the experimental results show that the model can effectively improve recognition accuracy.