Predictable

What is it?

A framework for actuarial modelling.

Installation

pip install predictable

Quick start example

A model.py file will be used to house the modelling logic which will be applied to each modelpoint.

# import the library
from predictable import CashFlow, DiscountFactors, Model, StaticCashFlow

# Create new model instance
model = Model()

# Add a premium component
model.add_component(
    CashFlow(
        input_array=[100], formula=lambda prev: prev * 1.05, label="premium"
    )
)

# Add a sum assured component
model.add_component(CashFlow(label="cover", input_array=[1_000_000]))

# Add an expense component
model.add_component(
    StaticCashFlow(
        input_array=[10, 10, 10, 10, 10],
        label="expense",
    )
)

# Add discounting component
model.add_component(DiscountFactors(interest_rate=0.05, label="V"))

# Project cashflows over term
# Results return a pandas df object
df = model.project(term=10)

# Perform linear combination style manipulations
# Discounting the components
components = ["premium", "cover", "expense"]
for component in components:
    df[f"V_{component}"] = df[component] * df["V"]


# Define reserving relationship
df["Reserve"] = df["V_cover"] + df["V_expense"] - df["V_premium"]

# Results get returned as a pandas dataframe
print(df)

Documentation

This project is documented using sphinx and the full documentation can be found at predictable.readthedocs.io.

Development & Contibutions

The following steps can be followed to set up a development environment.

1. Clone the project:

   git clone https://github.com/RatulMaharaj/predictable.git
   cd predictable

2. Install hatch

   pipx install hatch

3. Enter the default environment (this will activate the default virtual environment and install the project in editable mode).

   hatch shell default

Testing

This project uses pytest for testing purposes. The tests can be found in the tests directory. Tests will run after every commit (locally) and on every push (using github actions) but can also be run manually using:

hatch run test

Linting

This project is linted using ruff and formatted with black. The linting and formatting can be run manually using:

hatch run lint

hatch run format

Editing the docs

The documentation for this project can be found in the docs directory. The documentation is created using mkdocs and can be viewed locally using:

hatch run docs:serve

The docs can also be built for deployment using:

hatch run docs:build

License

MIT