Getting Started

Getting Started

Get up and running with Private Credit in 5 minutes.

Installation

1

pip install privatecredit

Or install from source:

1
2
3

git clone https://github.com/Digital-AI-Finance/private-credit.git
cd private-credit
pip install -e .

Quick Example

1. Generate Synthetic Data

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

from privatecredit.data import LoanTapeGenerator, MacroScenarioGenerator

# Generate loan portfolio
generator = LoanTapeGenerator(
    n_loans=10000,
    n_months=60,
    n_vintages=24,
    random_seed=42
)

# Generate loan tape and monthly panel
loans_df, panel_df = generator.generate()

print(f"Generated {len(loans_df)} loans")
print(f"Asset class distribution:")
print(loans_df['asset_class'].value_counts())

Output:

1
2
3
4
5
6

Generated 10000 loans
Asset class distribution:
corporate      4000
consumer       2500
realestate     2500
receivables    1000

2. Generate Macro Scenarios

1
2
3
4
5
6
7
8
9

# Generate macro scenarios
macro_gen = MacroScenarioGenerator(n_months=60)

baseline = macro_gen.generate_scenario('baseline')
adverse = macro_gen.generate_scenario('adverse')
severe = macro_gen.generate_scenario('severely_adverse')

print(f"Baseline GDP growth: {baseline['gdp_growth_yoy'].mean():.2%}")
print(f"Adverse unemployment: {adverse['unemployment_rate'].max():.2%}")

3. Run Baseline Model

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

from privatecredit.models import MarkovTransitionModel, PortfolioLossModel

# Create Markov model
markov = MarkovTransitionModel(asset_class='corporate')

# Compute PD curve
pd_curve = markov.compute_pd_curve(horizon_months=60)
print(f"12-month PD: {pd_curve[11]:.2%}")
print(f"Lifetime PD: {pd_curve[-1]:.2%}")

# Simulate portfolio loss
loss_model = PortfolioLossModel(markov)
results = loss_model.simulate_portfolio_loss(
    loans_df.head(1000),
    baseline,
    n_simulations=500
)

print(f"Expected Loss: {results['expected_loss_pct']:.2%}")
print(f"VaR 99%: EUR {results['var_99']:,.0f}")

Next Steps

  1. Tutorials - Step-by-step guides for each component
  2. Architecture - Understand the model hierarchy
  3. API Reference - Complete API documentation

Example Notebooks

Interactive Jupyter notebooks are available in the notebooks/ directory:

  • 01_data_exploration.ipynb - Explore synthetic data
  • 02_model_training.ipynb - Train deep generative models
  • 03_scenario_analysis.ipynb - Stress testing and scenarios

Back to Home