heavylight 1.0.1

Heavylight Actuarial Modelling Framework

heavylight

A lightweight actuarial modelling framework for Python

• single script
• installation optional: package with your models.
• only depends on pandas

Components

Model: - projection controller - class to subclass with your proprietary models - BeforeRun and AfterRun methods

Table: - simple long format table object - type information encoded via |int, |float, |str header suffixes

Usage

Create your model as a subclass of Model. Each model variable is defined as a method:

class Annuity(Model):
    def t(self, t):
        return t

    def expected_claim(self, t):
        return self.number_alive(t) * self.data["annuity_per_period"]

    def number_alive(self, t):
        if t == 0:
            return self.data["initial_policies"]
        else:
            return self.number_alive(t - 1) - self.deaths(t - 1)
    
    def deaths(self, t):
        return self.number_alive(t) * self.mortality_rate(t)

    def mortality_rate(self, t):
        return 0.02

Define input data as a dictionary

policy_data = {
    "initial_policies": 10,
    "annuity_per_period": 55,
}

Call the model, passing in the data dictionary, with a projection length of 20.

model = Annuity(data = policy_data,
                do_run = True,
                proj_len = 20,
                )

display result as a pandas table

model_cashflows = model.ToDataFrame()

Use numpy_financial or another package to calculate NPVs.

Notes

Projecting vs. discounting

• This package is designed for projecting actuarial variables, and calculates t=0, 1... in order.

• Actuarial models are generally highly recursive.

• If you create a method which refers to future t value (such as an NPV function) you may hit the python stack limit.

• The recommended solution is to project forward first, and then calculate T0 metrics based on the result.