The Influencial Factors Analysis And Short Term Forecasting On The China Containerized Freight Index

China Containerized Freight Index(CCFI)is the only index that reflects the containerized freight rate in the world.This index not only objectively and timely reflects the overall level of China’s containerized freight rate,but also quantifies the changes of China’s container shipping market.It has an important impact on the transportation cost of trade and is the "barometer" of China’s shipping market.The analysis and prediction of the influencing factors of the index is helpful for the shipping industry and the government to grasp the dynamics of containerized freight rate in time,avoid shipping risks,and improve the cost management level of the shipping industry.Based on the improved overall average empirical mode decomposition algorithm(Modified CEEMD,MEEMD)algorithm,this paper first decomposes the CCFI index to obtain 5 IMF components and a residual term,and then calculates the complexity of each component according to the fuzzy entropy algorithm for classification and combination,and reconstructs the CCFI index into high-frequency terms and lowfrequency terms and trend terms.Next,by reading a large amount of literature,combining theoretical and practical analysis,we selects 11 variable indicators that may have an impact on the changes in the CCFI index.The Granger causality test was performed between the high-frequency terms,low-frequency terms,and trend terms of the CCFI index and the selected influencing factor.Then use the ARIMAX model,BP neural network,MEEMD-ARIMAX model and MEEMD-BP model to predict the CCFI index,and evaluate the prediction effect of each model respectively.The results show that the Granger causes of CCFI include WTI,PMI,money supply and industrial producer price index.The Granger cause of high-frequency terms of CCFI include WTI and Consumer Price Index.The Granger causes of low-frequency terms of CCFI include WTI,first-order difference sequence of exchange rate,PMI,money supply and industrial producer price index,first-order difference sequence of unemployment rate,Shanghai composite index,macroeconomic prosperity index and export merchandise price index.The Granger cause of CCFI’s trend term is import merchandise price index.The prediction results show that traditional forecasting models when combine with MEEMD method can improve prediction performance.Through comparison,we found that the prediction performance of the MEEMDARIMAX combination model is the best among benchmark models.