Research On BDI Index Prediction Model Based On Key Characteristics

BDI index is the main freight index of international marine market,and it is also the marine market development barometer.To analyze and grasp the key characteristics of BDI index is of great significance for the prediction of BDI index,more in-depth understanding of the law of marine freight,and assistance in marine risk management.Centre on the above research purposes,the BDI index cycle,mean reversion and power-law characteristics are profoundly analyzed,by using the daily/weekly/monthly data series of BDI index,which has been set up since 1985 to the end of 2015.Furthermore,the three key characteristics of BDI index are applied to prediction.The main conclusions are as follows:1.By means of Variational Modal Decomposition(VMD),Grey Correlation Clustering(GRGC)and Fast Fourier transform(FFT)with estimates of the BDI index periodicity characteristics.The empirical results show that the BDI index long period about 16 years,middle period about 2-4 years,short period about 0.4-0.6 years.During the period of 2016-2020,the BDI index will remain in the long period of the depression stage and the middle period of the recovery stage.2.By using the three Self-excitation Regression model(3R-SETAR),an in-depth analysis of the BDI index nonlinear mean reversion characteristics.The empirical results show that the BDI index adjustment process is a three regimes nonlinear mean reversion process,which existing the marginal logarithmic growth zone region.The daily logarithmic growth rate is mainly in the internal regime and the weekly/monthly logarithmic growth rate is in the high regime.The internal regime volatility of BDI index is higher than the low and high regimes.Meanwhile,the recovery period in the low regime is shorter,and the recovery period is longer.3.By using Pareto,Exponential,Fokker-Planck and Gamma function,the power law characteristics of BDI index jump time and jump range are analyzed.The empirical results show that the BDI index logarithmic growth rate distribution has a peak of thin tail and volatility clustering.Gamma function fitting jump time is more appropriate,and Exponential function fitting jump BDI index is more appropriate.BDI index jump time and jump range are thin tailed power-law characteristics,the upper and lower are symmetry.4.A new prediction model of BDI index based on O-U stochastic process is constructed by combining periodicity,mean reversion and power law jump characteristics.Meanwhile,BDI index linked ship finance leasing range accrual interest swap is set up.The BDI index daily data of 2013-2015 is fitted by O-U stochastic model,and the BDI index in the first half of the year 2016 is predicted.The case shows that the model has higher prediction accuracy.BDI index linked ship finance leasing range accrual interest swap example analysis is a good explanation of practical problems.