MarineBees

Our Approach

Not just a score, but the reasoning behind it

A risk score is only as valuable as its ability to explain why it has that value. We build our models on maritime expertise and strengthen them with data.

SIRE 2.0

The inspection has changed, and so has the data

With SIRE 2.0, the inspection has moved from a checklist to an assessment focused on context and human behaviour. This gives a more realistic picture, but it also makes the data harder to analyse with traditional methods.

01

Dynamic question set

Questions change according to the vessel's type, cargo and operation.

02

Inspection context

The same finding can mean different things in different operations.

03

Human factors

The crew's behaviour and decision-making are also observed.

04

Evidence-based verification

Findings are supported by photos, documents and observations.

05

Open-ended assessment

Inspector comments are recorded as free text.

Hybrid model

Where expert knowledge meets data

Models trained on data alone lose reliability when data is scarce and variable. That is why we use a two-stage structure.

1

Expert-based risk model

Cause-and-effect relationships between findings and risk outcomes are built on maritime expertise.

Human factorsTechnical systemsOperational processes
2

Calibration with data

The model is trained on historical inspection data and updates itself as new data arrives.

Machine learningNatural language processingContinuous learning
=

Explainable result

Every risk score comes with its cause, contributing factors and suggested actions.

Risk scoreContributing factorsSuggested action

How this approach shows up in the product

See how the hybrid model is used in the inspection process on the SIRE 2.0 Self-Assessment page.