A novel short-time ship traffic flow prediction model was proposed to meet the needs of accurate and rapid prediction of short-time ship traffic flow. The proposed model optimizes Long Short Term Memory (LSTM) neural network based on Ensemble Empirical Mode Decomposition (EEMD) and Permutation Entropy (PE). The prediction process was divided into three stages. Firstly, EEMD was used to reduce the non-stationarity effect of short-term ship traffic flow. Secondly, PE was used to reconstruct the phase space of the time series components decomposed by EEMD to greatly shorten the operation time of the prediction model. Finally, the reconstructed high and low frequency components and residuals are imported into the LSTM model for prediction, and the final prediction result can be obtained after the predicted values are superimposed. The effectiveness of the model is verified by collecting 4392 sets of ship traffic flow data from the main channel of Qingdao Port for 366 consecutive days. The results show that the prediction performance of the proposed EEMD-PE-LSTM model is the best, which is improved by 167% and 226% re-spectively. The proposed model was proved that can reflect the changes of the future short-term ship traffic flow more accurately and quickly. Furthermore, the K-means clustering algorithm was used to cluster the prediction results, which can present the dy-namic transformation process of the ship traffic state more intuitively and simplify the difficulty of manual identification. The visu-alization of the channel traffic state is realized and the workload of front-line ship traffic management personnel is effectively re-duced using the clustering process.