A study on ship collision conflict prediction in the Taiwan Strait using the EMD-based LSSVM method

Area of Science:

• Maritime Safety
• Predictive Analytics
• Ocean Engineering

Background:

• Ship collision accidents pose significant risks to maritime traffic safety, causing casualties and environmental damage.
• Assessing regional collision risk is crucial for navigators and surveillance, but historical data presents limitations.
• Ship collision conflicts serve as a viable proxy for analyzing maritime traffic safety and developing countermeasures.

Purpose of the Study:

• To propose and validate a hybrid model for accurate forecasting of ship collision conflicts.
• To enhance the quantitative study of maritime traffic safety problems.
• To provide a reliable tool for predicting potential maritime incidents.

Main Methods:

• A hybrid Empirical Mode Decomposition (EMD) and Quantum-behaved Particle Swarm Optimization (QPSO) optimized Least Squares Support Vector Machine (LSSVM) model was developed.
• Time series data of ship collision conflicts were decomposed using EMD into intrinsic mode functions (IMFs) and a residue.
• QPSO algorithm optimized LSSVM parameters for forecasting each subseries, with final predictions obtained by summing individual forecasts.

Main Results:

• The proposed EMD-QPSO-LSSVM approach demonstrated high efficiency in ship collision conflict prediction.
• Comparative analysis showed the method outperformed traditional models like GM and Lasso regression, as well as EMD-ENN.
• The model effectively captures the complex dynamics within ship collision conflict data.

Conclusions:

• The EMD-QPSO-LSSVM model offers a robust and efficient solution for predicting ship collision conflicts.
• This predictive capability can significantly contribute to improving maritime traffic safety and risk management.
• The study highlights the potential of hybrid intelligent models in addressing complex challenges in maritime operations.