Rate with a rating model - Effective

A rating model computes one part of a carrier’s rating plan, such as a credit score, a tier or a tier factor. Install Relativities and set your API key before running this page.

List the models

from relativities import RatingModels

models = RatingModels.list(layer="factor")
sorted(models)  # ['example.mileage-factor']

RatingModels.list returns the newest upload of each model in your organization’s catalog. Filter by carrier, layer, state or fidelity. Each layer is score, tier or factor.

Pick a model and check its sources

model = models.example.mileage_factor
model.descriptor["fidelity"]  # 'exact_filed'
model.descriptor["sources"]  # [{'serffTrackingNumber': 'EXMP-000000001', 'pages': 'pp.4-5'}]

The part of the model ID before the first . groups the models by carrier. The rest of the ID becomes the attribute, with each - and . written as _. models["example.mileage-factor"] names the same model. The descriptor records the filings behind the model and its knownGaps. Its fidelity says how closely the model follows them:

exact_filed: the filed tables and rules, as filed.
alt_state_proxy: one state’s filing stands in for another state’s confidential one.
non_filed_assumption: a step that rests on an assumption, with the assumption in knownGaps.

Rate inputs

submission = {
    "Policy": {
        "drivers": [{"age": 42}],
        "vehicles": [{"annual_mileage": 5000}],
        "coverages": {"collision": False},
    }
}
results = model.rate(submission)
results.to_pandas()["annual_premium"].tolist()  # [400.0]

The model downloads and verifies its bundle the first time it rates. Rating then runs on your machine, on the same native engine as a Rater. rate and explain take the model’s own inputs, as model.input_schema describes them.

Pin an upload

from relativities import RatingModel

pinned = RatingModel.load("example.mileage-factor", bundle_id=model.bundle_id)
pinned.bundle_id == model.bundle_id  # True
pinned.close()
model.close()

Rating models use immutable uploads in place of releases. Each upload’s bundle_id names it exactly. Pass bundle_id to rate with one upload every time. After the first download, local_files_only=True loads that upload from the local cache, offline.