A multi-target tracking model based on the Hungarian algorithm and sliding window Convolutional Long short term memory neural network (SW-CLSTM) is proposed to study the anonymous position information of enemy ship formation in the past. The model training process is data-driven and aims to achieve accurate target association and accurate prediction through the effective use of data. Firstly, the Hungarian algorithm is used to conduct effective association between the track data, and then sliding window technology is used to screen and preprocess the data, which is then input into the convolutional long short-term memory neural network. Through the output of dynamic sliding window, the model can accurately predict the ship's future navigation trajectory, so as to achieve efficient multi-target tracking function. The experimental part firstly verifies the superiority of the Hungarian algorithm for data association. Secondly, the SW-CLSTM model and the LSTM model are compared respectively in three scenarios of linear navigation, turning and continuous turning. The results show that this model has better performance than the comparison model.