The invention discloses a ship collision risk evaluation method and system based on data mining. The method includes: according to recorded ship collision historical data, predicting causation factor occurring times through a cumulative auto-regressive moving average model, adopting a principal component analysis method to perform dimension reducing calculation on causation factors, and taking an acquired principal component as input of a first neural network to predict collision probability; adopting decision-making tree analysis to find an accident sequence with high confidence degree and occurring frequency, using a neural network model to calculate number of injured people, number of dead people and property loss, and performing standardization and normalization processing to eliminate influence of order and dimension; using an expert system and a fuzzy hierarchy analysis method to give different weights to indexes to quantify collision consequence. According to the above content, the ship collision risk evaluation system based on data mining is developed, ship collision risk is predicted quantitatively, real-time guidance is provided for ship control, collision avoiding emergency and on-site handling, and loss can be lowered effectively.