This paper presents a data mining-then-predict method for proactive maritime traffic management using machine learning. The method leverages Automatic Identification System (AIS) data to extract and predict traffic patterns in inland waterways. The approach involves data processing, data mining, and traffic pattern prediction. Data processing includes cleaning AIS data, trajectory reconstruction, and matching static and dynamic data. Data mining identifies traffic groups based on origin and destination zones, while traffic pattern prediction uses a decision tree model to classify ship trajectories into specific patterns. The method was validated in the convergent areas of the Yangtze River and the Hanjiang River, achieving a 96.7% accuracy rate in predicting single ship trajectories and 80.9% precision. The method effectively extracts and predicts traffic patterns, identifies congestion in convergent waters, and supports traffic management strategies. The study also explores the impact of ship sailing time on traffic pattern prediction accuracy. The proposed method is highly interpretable and provides a framework for inland maritime traffic situation awareness and prediction research. The results demonstrate the effectiveness of the method in enhancing maritime traffic management through proactive traffic pattern analysis and prediction.This paper presents a data mining-then-predict method for proactive maritime traffic management using machine learning. The method leverages Automatic Identification System (AIS) data to extract and predict traffic patterns in inland waterways. The approach involves data processing, data mining, and traffic pattern prediction. Data processing includes cleaning AIS data, trajectory reconstruction, and matching static and dynamic data. Data mining identifies traffic groups based on origin and destination zones, while traffic pattern prediction uses a decision tree model to classify ship trajectories into specific patterns. The method was validated in the convergent areas of the Yangtze River and the Hanjiang River, achieving a 96.7% accuracy rate in predicting single ship trajectories and 80.9% precision. The method effectively extracts and predicts traffic patterns, identifies congestion in convergent waters, and supports traffic management strategies. The study also explores the impact of ship sailing time on traffic pattern prediction accuracy. The proposed method is highly interpretable and provides a framework for inland maritime traffic situation awareness and prediction research. The results demonstrate the effectiveness of the method in enhancing maritime traffic management through proactive traffic pattern analysis and prediction.