Research on Ship Trajectory Prediction Based on LSTM Algorithm

Precise prediction of vessel movements is becoming ever more crucial for traffic management and collision prevention in maritime environments as worldwide shipping continues to expand. Precise prediction of vessel movements is becoming ever more crucial for traffic management and collision prevention in maritime environments as worldwide shipping continues to expand. This study presents an LSTM-based prediction framework designed to address the nonlinear characteristics of vessel motion and the irregularities commonly found in Automatic Identification System (AIS) data, including uneven sampling intervals and missing observations. A structured preprocessing pipeline is applied, using linear interpolation to achieve temporal consistency and min–max normalization to eliminate scale disparities across features. The proposed deep learning architecture integrates LSTM and dense layers to forecast vessel latitude, longitude, speed, and heading. Experimental results demonstrate that the optimized model attains high predictive performance, achieving an average single-step error of 16.5 m and an accuracy of 98.84%. Furthermore, the model exhibits strong generalization in multi-step prediction tasks, successfully capturing dynamic navigational behaviors such as turning patterns. These findings indicate that the proposed method provides a reliable and effective solution for vessel route prediction, offering practical value for maritime monitoring, route planning, and navigational safety applications.