BDI & SCFI prediction (‘22/14W)
Firstly, the BDI and SCFI values predicted last week and the average actual BDI and SCFI values of this week are compared and tested.
1. Last week's forecasts
1-1. BDI, Actual Vs. Predicted
| Date | Last week Forecast | Actual BDI | MAPE(%) | Accuracy(%) |
| ‘22/14W | 2,483.531 | 2,397.00 | 3.6% | 96.4% |
1-2. SCFI, Actual Vs. Predicted
| Date | Last week Forecast | Actual SCFI | MAPE(%) | Accuracy(%) |
| ‘22/14W | 4,430.827 | 4,348.81 | 1.9% | 98.1% |
The prediction accuracy of BDI and SCFI was 96.4% and 98.1%, respectively. Although it is a very good number in terms of prediction accuracy, the purpose of this study is to understand the trend flow of BDI and SCFI. Even if the trends and trends are right, it will be of great help in judging the market conditions.
2. BDI Index Outlook
To forecast the BDI index, we used weekly data (426 pieces) from January 3, 2014 to April 1, 2022 to predict the fare for three weeks in the future.
| Date | Forecast | Trend(WoW) |
| ‘22/15W | 2244.468 | ↓ |
| ‘22/16W | 2163.702 | ↓ |
| ‘22/17W | 2155.168 | → |
As shown in the table and figure above, the BDI (as of April 1st: 2,397.00) is expected to decline significantly next week. In the 15th week, the forecast value is 2,244.468, which is -6.4% lower than that of 4/1, and it is expected to drop to -3.6% at 16W compared to 15W. Entering 17W, the decline is stable to -0.4% and is expected to remain flat.
3. SCFI Outlook
For the SCFI index forecast, we used weekly data (624 pieces) from October 16, 2009 to April 1, 2022 to predict index for three weeks in the future.
| Date | Forecast | Trend(WoW) |
| ‘22/15W | 4331.830 | ↓ |
| ‘22/16W | 4334.999 | → |
| ‘22/17W | 4345.949 | ↑ |
As shown in the table and figure above, the SCFI (as of April 1st: 4,348.71) is expected to gradually decrease the decline. At the 15th week, the forecast value is 4,331.830, which is -0.4% lower than on 4/1, but it is expected to stop falling and turn upward at 16W. And it is predicted that 17W will increase by 0.3% compared to 16W.
I want to test the predicted values through continuous analysis. In addition to ARIMA, which is currently being used for analysis, other econometric, machine learning, and deep learning analysis will also be utilized.
thank you.