microsoft/qlib

Features

• Algorithmic Trading Frameworks - Provides a modular architecture for building and simulating automated investment logic from portfolio management to order execution.
• Algorithmic Trading Platforms - Coordinates multiple trading strategies and monitors their performance through historical simulation and real-time execution.
• Algorithmic Trading Simulators - Simulates historical market performance to evaluate the effectiveness of investment strategies before deployment.
• Frameworks - Facilitates the development of data-driven investment models through specialized financial data processing and predictive algorithm training.
• Research Platforms - Provides an end-to-end pipeline for developing and testing financial trading strategies through data ingestion and model training.
• Financial Machine Learning Models - Applies predictive modeling and reinforcement learning to large-scale market datasets to automate investment decisions.
• Financial Machine Learning Toolkits - Provides a specialized collection of tools for training predictive models and reinforcement learning agents on historical market data.
• Trading Strategy Definitions - Enables the definition of custom logic for market predictions and automated trading decisions.
• Algorithmic Trading - Defines custom portfolio management logic by extending base classes to generate specific trade decisions.
• Financial Forecasting Models - Calculates stock scores using a modular interface that integrates with automated workflows.
• Automated Trading Execution - Executes pre-built investment strategies to manage stock holdings and generate order lists.
• High-Performance Data Infrastructures - Maximizes throughput for large-scale financial datasets while ensuring point-in-time data integrity.
• Quantitative Workflow Orchestrators - Manages complex research pipelines through configuration-driven automation to ensure reproducible experiments.
• Workflow Orchestrators - Automates complex research pipelines including data processing, model training, and backtesting via a centralized configuration interface.
• Reinforcement Learning Algorithms - Optimizes long-term trading performance by defining agents, environments, policies, and reward signals.
• Forecasting Models - Data distribution generation for concept drift adaptation.
• Machine Learning - AI-oriented quantitative investment and research platform.
• Machine Learning Frameworks - AI-oriented platform for quantitative investment research.
• Financial Analysis - AI-oriented platform for quantitative investment research.
• Quantitative Frameworks - AI-driven quantitative investment framework for research and production.
• Trading and Backtesting - AI-oriented quantitative investment platform.
• Centralized Data Management Systems - Stores and accesses information from a central server that uses advanced caching techniques to speed up calculations.
• Configuration-Driven Orchestrators - Automates complex research pipelines and experiment tracking through declarative configuration files and standardized interfaces.
• Pipeline Execution Engines - Runs complete analytical workflows including data preparation and model training via single-command execution.
• Backtesting Engines - Prevents data leakage during backtesting by ensuring models only access information available at specific historical timestamps.
• Machine Learning Experiment Trackers - Tracks metrics, parameters, and tags for machine learning runs using a hierarchical system.
• Meta Learning Frameworks - Learns patterns across multiple forecasting tasks to apply acquired knowledge toward guiding future model training.
• Reinforcement Learning Environments - Implements simulator, state interpreter, action interpreter, and reward function components for reinforcement learning.
• Reinforcement Learning Training Pipelines - Coordinates the reinforcement learning lifecycle by managing environment creation, policy updates, and runtime execution.
• Temporal Data Management - Enforces strict historical boundaries to prevent future information leakage during backtesting and model training.
• Experiment Tracking Systems - Organizes research projects by tracking individual experiment records and comparing configurations to evaluate progress.
• Automated Experimentation Tools - Generates multiple research tasks from templates by varying parameters to facilitate automated model testing.
• Model Integration Interfaces - Connects custom forecasting models by extending base classes and configuring parameters via external files.
• Reinforcement Learning Strategies - Develops quantitative trading strategies using reinforcement learning by following structured learning paths.
• Binary Data Formats - Provides custom binary file structures to optimize disk I/O and memory throughput for large-scale market datasets.
• Technical Analysis - Provides tools for building mathematical expressions using historical market data to predict future asset returns.
• Configuration Management Systems - Instantiates research components like models and data handlers using external configuration files to reduce boilerplate code.
• Performance Analysis - Generates visual summaries and performance metrics for investment portfolios to evaluate strategy effectiveness.

Star history