Predictive risk profiling of vessels

Seven Ocean Research Pty Ltd runs a weekly risk prediction routine at ship level (for each IMO) for the world fleet (all vessels above 100gt with IMO) based on statistical models and machine learning (ML). 

The models are based on a comprehensive global dataset with over 600 evaluated risk factors. 

Available endpoints of interest at ship level:

• 9 incident type models predicting incidents with very serious or serious consequences (eg. total loss, fire and explosion, loss of life, pollution, machinery failures)

• 13 detention and deficiency type models predicting detention or finding specific deficiencies

• 4 damage type models: predicting the monetary value at risk (MVR) and conditional damage type models: MVR is a proxy to incident consequences of the following damage types: 1) Hull and Machinery, 2) Loss and Life and Injuries, 3) Pollution and 4) Other Marine Liabilities. MVR is also split into an upper and lower boundary to indicate uncertainty.

• combination models

Output format: 

• weekly or monthly csv file for all vessels above 100gt (around 135k vessels)

• percentile ranks, risk categories or probabilites at ship level (IMO level)

Selected Publications

Maritime Risk Assessment for Portugal, Sweden and the Baltic Sea with an emphasis on shadow fleets