Price Forecasting of Marine Fish Based on Weight Allocation Intelligent Combinatorial Modelling

Area of Science:

• Fisheries Science
• Econometrics
• Artificial Intelligence

Background:

• Marine fish markets exhibit complexity and uncertainty, challenging traditional price forecasting.
• Accurate price prediction is vital for socio-economic development and the fisheries industry.
• Existing forecasting methods struggle with marine fish market volatility.

Purpose of the Study:

• To enhance the accuracy of marine fish price prediction using an intelligent combination model.
• To address the limitations of traditional forecasting methods in volatile markets.
• To develop a robust model for predicting marine fish prices.

Main Methods:

• Decomposition of price series using empirical wavelet transform, singular spectrum analysis, and variational mode decomposition.
• Cross-prediction on decomposed series using bidirectional long short-term memory (BiLSTM), extreme learning machine (ELM), and exponential smoothing.
• Weight allocation and combined prediction using the Particle Swarm Optimization-Chaser-Stalker (PSO-CS) intelligence algorithm.

Main Results:

• The intelligent combination model with PSO-CS weight allocation achieved higher prediction accuracy than single models.
• Empirical analysis using daily sea purchase price data for Larimichthys crocea demonstrated superior performance.
• The model showed enhanced stability in predicting marine fish prices.

Conclusions:

• Intelligent combinatorial modeling with weight allocation improves marine fish price prediction accuracy and stability.
• The developed model effectively adapts to market changes and price fluctuations.
• This approach offers a significant advancement for fisheries industry forecasting.