CCFI Index Prediction Based On Variational Mode Decomposition And Neural Network

As global trade continues to develop,the position of the container shipping market in the international shipping market is becoming increasingly important.Due to the influence of various factors such as global economy,policies and regulations,and war conflicts,the volatility of the container shipping market is also becoming more and more evident.This volatility not only has a huge impact on the development of economic trade but also increases the investment risks of shipping companies.China Container Freight Index(CCFI),Baltic Dry Index(BDI),and Baltic Dirty Tanker Index(BDTI)are collectively known as the world’s three major freight indices.The CCFI index objectively reflects the changes in the Chinese and global shipping markets to a certain extent.The CCFI index can not only serve as a leading indicator for measuring international economic and trade exchanges but also has an important impact on the transportation costs of goods trade.Therefore,building a scientific and effective CCFI index prediction model can help the shipping industry and government timely understand the development trend of the container shipping market,formulate corresponding response strategies,minimize risks,and improve market competitiveness.Based on the aforementioned requirements,this thesis proposes a CCFI index prediction model based on variational mode decomposition(VMD)and neural network.The model first uses the VMD method to denoise the time series of CCFI index,and then combines convolutional neural network(CNN)and long short-term memory(LSTM)network for combined prediction,in order to solve the shortcomings of traditional prediction methods in prediction accuracy.The main work of this thesis is divided into data decomposition and data prediction.Firstly,the CCFI index data of the past 20 years is selected,and VMD decomposition is used to denoise the original data to obtain K intrinsic mode functions.The applicability of VMD decomposition is compared and verified with empirical mode decomposition(EMD)decomposition.Then,the data reconstructed by VMD decomposition is used as the prediction sample,and short-term features are extracted by the CNN network,and the feature vector is input into the LSTM network for training and testing in a 7:3 ratio.The EMD-LSTM(EL)model,VMD-LSTM(VL)model,EMD-CNN-LSTM(ECL)model,CNNLSTM(CL)model,and LSTM model are compared and tested.Through the analysis of the prediction results,the superiority of the proposed model is validated.Finally,the CCFI index is predicted for the next year.The research results indicate that:(1)this article uses a network model combining variational mode decomposition for index processing and neural network for index prediction,which can effectively predict the CCFI index.The evaluation indicators including MAPE,RMSE,and MAE show that the model proposed in this article is significantly superior to other comparative models in terms of prediction accuracy and stability.It provides a new method and idea for freight rate prediction and has certain reference value.(2)Based on this,the CCFI index for the next year is predicted,and the result shows that the container shipping market freight rates may first decline and then rise in 2023,with a much greater degree of decline than the rate of increase and a relatively slow pace of rise.This means that the CCFI freight rate may remain at a low level for a period of time,and it is expected to gradually recover with the gradual recovery of the market.