A novel self-adaptive multivariate grey model with external intervention for port cargo throughput prediction

Purpose Accurate prediction of port cargo throughput within Free Trade Zones (FTZs) can optimize resource allocation, reduce environmental pollution, enhance economic benefits and promote sustainable transportation development. Design/methodology/approach This paper introduces a novel self-adaptive grey multivariate prediction modeling framework (FARDCGM(1,N)) to forecast port cargo throughput in China, addressing the challenges posed by mutations and time lag characteristics of time series data. The model explores policy-driven mechanisms and autoregressive time lag terms, incorporating policy dummy variables to capture deviations in system development trends. The inclusion of autoregressive time lag terms enhances the model’s ability to describe the evolving system complexity. Additionally, the fractional-order accumulative generation operation effectively captures data features, while the Grey Wolf Optimization algorithm determines optimal nonlinear parameters, enhancing the model’s robustness. Findings Verification using port cargo throughput forecasts for FTZs in Shanghai, Guangdong and Zhejiang provinces demonstrates the FARDCGM(1,N) model’s remarkable accuracy and stability. This innovative model proves to be an excellent forecasting tool for systematically analyzing port cargo throughput under external interventions and time lag effects. Originality/value A novel self-adaptive grey multivariate modeling framework, FARDCGM(1,N), is introduced for accurately predicting port cargo throughput, considering policy-driven impacts and autoregressive time-lag effects. The model incorporates the GWO algorithm for optimal parameter selection, enhancing adaptability to sudden changes. It explores the dual role of policy variables in influencing system trends and the impact of time lag on dynamic response rates, improving the model’s complexity handling.