We support four types of datasets, each one has a separate endpoint and requires the data to be in .json format.
- Sensor data
- Noon reports
- Event data
- Weather data: not required. In case it is not present, we fetch it from a 3rd party source.
Useful frequencies for sensor data would be somewhere between 1 second and 1 hour, ideally it is 5 minutes.
The required time range for a useful model in practice starts at 3 months of data while at sea. We recommend to deliver as much as is available, preferably several years. This allows us to tell a more complete story on how performance evolves over time.
The following variables should be present in either the noon reports and/or sensor data:
|must include timezone
|Location: Latitude + Longitude
|Speed Over Ground or Speed Through Water
|Main Engine RPM
|Main Engine Power
|Main Engine Fuel Oil Consumption
|Usually more reliable in noon reports
To speed up the modelling process and ensure the highest model quality we highly recommend uploading as many variables as you have available. So if Draft is available in both noon reports and sensor data available, upload both. That way our systems can use the most stable one.
We also need to know when ship performance-altering events (hull cleanings, dry-dockings, ESDs, ...) occurred during the period for which any data is given. Only providing the correct labels (
esd ...) with a timestamp is sufficient, but any additional information can help.
In case any other relevant performance events have occurred, use the
We impose minimal checks on each data sample.
- Each sample should at least have a timestamp in ISO8602 format.
- Each parameter should have the correct data type. For example, text won't be accepted where a number should be.
- Parameters may be missing or null.
Data can be corrected by re-uploading it with the same timestamp. We will always use the values of the latest uploaded sample.
In case erroneous timestamps were uploaded, contact us to fix the issue.
Updated 24 days ago