Work

gymfolio: A Reinforcement learning environment for Portfolio Optimization in Python

Year: 2025

Type: software-paper

Abstract: This paper introduces gymfolio , a modular and flexible framework for portfolio optimization using reinforcement learning. gymfolio is built around the PortfolioOptimizationEnv class, enabling seamless integration of market observations, technical indicators, and dynamic rebalancing strategies. The implementation emphasizes adaptability, supporting extensions for custom investment goals and trading scenarios. gymfolio serves as a lightweight adaptable environment for advancing research in financial machine learning and promoting innovation in portfolio strategy development.

Source: SoftwareX

DOI: 10.1016/j.softx.2025.102106

Authors: Francisco Espiga-Fernández, Álvaro García-Sánchez, Joaquín Ordieres-Meré

Institution: Universidad Politécnica de Madrid

Language: English

Topic: Stock Market Forecasting Methods

Subfield: Management Science and Operations Research

Field: Decision Sciences

Domain: Social Sciences

Open Access status: gold

Estimated APC paid: $1,200

Funders: Ministerio de Ciencia, Innovación y Universidades, Ministerio de Ciencia e Innovación, European Regional Development Fund, Agencia Estatal de Investigación