PyStatsInsurance

Actuarial and insurance statistical computing for Python.

Status: 0.1.0 — first module shipped. Chain-ladder reserving is available now; more actuarial methods are on the roadmap below.

PyStatsInsurance is part of the open-core PyStatistics family:

Like its siblings, it builds on pystatistics for the general statistical layer and adds methods specific to actuarial and insurance work.

What's available now (0.1.0)

Chain-ladder loss reserving from a cumulative-claims development triangle (unknown lower-right entries marked NaN):

import numpy as np
from pystatsinsurance import reserving

triangle = np.array([
    [100.0, 150.0, 180.0],
    [110.0, 165.0, np.nan],
    [120.0, np.nan, np.nan],
])
result = reserving.chain_ladder(triangle)
print(result.summary())

chain_ladder computes volume-weighted age-to-age development factors, the completed triangle, ultimate claims per origin period, and the outstanding reserve per origin period and in total. It fails loud on a ragged / non-2×2 / negative triangle and when a development factor cannot be derived.

Roadmap (candidates, not commitments)

• Mack chain-ladder (reserve standard error) and Bornhuetter–Ferguson reserving.
• Frequency / severity loss models and aggregate-loss distributions.
• Credibility theory and life-actuarial methods.

Installation

pip install pystatsinsurance

License

MIT © Hai-Shuo. Part of the SGCX open-core ecosystem.