Forecasting China's shipping indices based on modal decomposition and optimized deep learning integrated model

Area of Science:

• Maritime Economics
• Financial Time Series Analysis
• Deep Learning

Background:

• Accurate forecasting of shipping indices is crucial for China's maritime sector.
• Traditional models struggle with the nonlinearity and non-stationarity of shipping data.

Purpose of the Study:

• To develop an innovative hybrid forecasting model, VMD-CPSO-BiLSTM, to enhance prediction accuracy for Chinese shipping indices.
• To address the challenges of nonlinearity, non-stationarity, and multi-scale characteristics in time series forecasting.

Main Methods:

• Variational Mode Decomposition (VMD) to decompose time series into intrinsic mode functions (IMFs).
• Chaotic Particle Swarm Optimization (CPSO) to optimize Bi-directional Long Short-Term Memory (BiLSTM) network parameters.
• Integration of predictions from high-frequency and low-frequency components for comprehensive forecasts.

Main Results:

• The VMD-CPSO-BiLSTM model demonstrated superior performance compared to conventional single deep learning models and other hybrid approaches.
• The model effectively captured nonlinear and complex patterns in shipping index data.
• Empirical validation using key Chinese shipping indices confirmed the model's enhanced predictive accuracy and stability.

Conclusions:

• The VMD-CPSO-BiLSTM model offers a reliable tool for forecasting shipping market trends.
• The model provides enhanced decision-making support for strategic planning and operational management in the maritime industry.
• This methodological innovation contributes significantly to maritime economics and financial time series analysis.