scikit-agent

A scikit-learn compatible toolkit for agent-based economic modeling.

scikit-agent is a scikit-learn compatible toolkit for agent-based economic modeling. It provides a unified interface for creating, solving, and simulating economic models using modern computational methods including reinforcement learning, neural networks, and traditional numerical techniques.

Key Features

  • Scikit-learn compatible API for easy integration with the Python scientific ecosystem

  • Economic model classes for consumption-savings, portfolio choice, and other standard models

  • Multiple solution algorithms including value function iteration, policy iteration, and neural network approaches

  • Simulation tools for generating synthetic data and running policy experiments

  • Comprehensive documentation with examples and tutorials

Quick Start

import skagent

# Create a basic consumption-savings model
model = skagent.models.ConsumptionSavingsModel(
    periods=50, discount_factor=0.95, risk_aversion=2.0
)

# Solve the model
model.fit()

# Run simulations
results = model.simulate(n_agents=1000, n_periods=50)

Installation

pip install scikit-agent

For development installation:

git clone https://github.com/scikit-agent/scikit-agent.git
cd scikit-agent
pip install -e ".[dev,docs]"

Indices and tables