Grenadier: L2-based Anomaly Detection
Overview
dxFeed Grenadier is an innovative tool designed to provide real-time inference of market anomalies. It empowers traders, investors, and financial institutions with the insights needed to understand market shifts and identify atypical order placing activities.
Grenadier leverages proprietary unsupervised machine learning models based on autoencoder architectures to detect early signs of potential regime changes, upcoming trade halts, or significant price movements. This service offers insights crucial for adaptive risk management, market making, or trade execution systems.
Grenadier is built on robust dxFeed technologies, including the Historical Data Lake (HDL) and dxLink, ensuring seamless integration, efficient training cycle on vast amounts of historical data, and optimal performance during online inference.
Core Features
Grenadier has been engineered for flexibility, offering real-time market insights to anticipate and react to significant price changes or disruptions, reducing risk and optimizing trade decisions.
Anomaly Score Developer API
Offers real-time assessments of market anomaly scores based on Level 2 order book data.
Reconstructed Order Book API (B2C Pro)
Provides a comprehensive view of the order book as interpreted by anomaly-free models.
Tailored ML Models (B2B)
Allows institutions and analysts to fine-tune models based on outlier preferences, with Pytorch compatibility.
Typical Use Cases
Early Market Shift Detection
Risk Management
Trade Execution
Сustom Model Integration
Novel Market Data Insights
Target Audience
dxFeed Grenadier is designed to cater to a diverse range of users, including:
Professional Traders
Gain competitive advantage by detecting anomalous activities in order data streams.
Institutional Investors
Enhance microstructural signals for risk management and trade execution strategies.
Quantitative Analysts
Utilize customized machine learning models for market analysis and proactive inference.
Financial Institutions
Integrate model outputs into existing data pipelines or fraud detection systems.
