For 300 years, Lloyd’s List intelligence have been the trusted independent partner and source of maritime data, insights, and expertise.
Managing the world’s largest maritime intelligence database, Lloyd’s List Intelligence supports 60,000+ professionals to evaluate and mitigate supply chain risk; analyze global trade and create operational efficiencies with first in class Data, Analytics and Platform solutions.
Here are just some of the data we are incorporating in Quantum ML’s data lake:
- Vessel movements & positions
- Container tracking
- Vessel Characteristics
- Vessel Ownership
- Vessel Casualties, Incidents and Sanctions
- Port data
- Financial and Credit data
- Oil and product cargo volumes
- Vessel risk indicators
The wealth of data that is available via Lloyd’s List is going to be a game changer for QDT and for the outcomes achieved with our machine learning solution. We have already demonstrated incredible results with this data and we have barely scratched the surface of assessing the full scope of it’s applicability in commodity price prediction, and other use-cases like freight and trade flow forecasting. The results from our initial data testing are summarized in the following section.
INITIAL DATA TESTING
- Target Variable: Australian Low Vol Coking Coal
- Prediction Timeframe: 1-40 days
- Data Prep: Data was transformed to produce both (1) Time at Port and (2) Number of Vessels at Port.
- Data Used in the Model: Total of 1428 total input predictors. Data categories include: Commodity price data, Fundamental data (Supply and Demand), Macroeconomics, News, Weather and Technical indicators.
- Lloyds Data: Lloyds Data in this test makes up 58 of the 1418 total data inputs in the test model.
Below you will see a summary of the most predictive/valuable data inputs. Additionally, we also show the same list filtered to show how the ports of interest rank amongst each other. The higher up on the list, the more it contributes to the signal, and therefore the more predictive it is.
“Feature importance” (the % column on the far left) indicates the relative contribution of a given predictor to the output signal of the model, measured over it’s total lifetime.
- You can see from the above that for the prediction of Coking Coal, Lloyds List shipping data ranks very high in terms of predictability, populating 14 of the top 40 spots. Other high ranking predictors are other grades of coking coal and Technicals indicators.
- Given how Lloyds data makes up only 58 out of 1418 data inputs in the model, this is a very strong indication of the value of this data in predicting PLV Coking Coal.
- These results are very encouraging and demonstrate Lloyd’s List shipping data to be the single most valuable data source amongst all of the data we have tested for coking coal.
- To put this in context, the other 1360 predictors represent an extensive list of commodities price data, fundamentals, macroeconomic, weather and news data, coming from 12 other data providers.
Keep in mind also that this test dataset also only incorporates data from bulk carriers for the specific ports of interest relevant to coking coal. This is really a small fraction of what Lloyd’s has in their database.
Needless to say, our team is very excited to keep exploring how this data will improve performance for other target variables. We have identified additional types of data which we anticipate will provide a significant boost to model performance. For example, for the prediction of Crude or Freight, we intend to expand the data to include other vessels classes as well as additional ports. Furthermore, Lloyd’s offers their comprehensive APEX database, which has an incredible breadth and depth of crude-related shipping data to utilize.