DataMine Machine Learning as a Service

Quantum Data Technologies (QDT) in partnership with CME Group, is excited to launch the DataMine Machine Learning Service, a transformative Machine Learning as a Service (MLaaS) solution. This state-of-the-art platform is designed to democratize access to sophisticated analytics, enabling users to generate actionable trading insights and make more informed investment decisions.

• Automate model generation
• Enhance trading portfolio
• Maximize resources
• Visualize trading strategies

Unique Data Provided by CME Group

CME Group supplies a wealth of proprietary data that forms the backbone of the DataMine ML Service, making it a standout offering in the financial services sector. This includes:

• Historical CME Price Data

‍End-of-day statistics, transactions, bids, and offers across various futures and options markets.

• Curated Exchange Data

‍Specialized data like option volatility metrics from Bantix, market sentiment analyses and other curated datasets that are essential for nuanced market analysis.

• Third-Party Data on CME Datamine

‍Access to external datasets such as RS Metrics, Predata, and Indigo Ag, among others, providing broader insights into global market trends and commodity health.

This unique blend of CME-specific and third-party data enables users to perform deep and diverse analyses, leveraging information that spans from macroeconomic indicators to specific commodity news, enhancing the predictive accuracy of their models.

Model Building Process

The model-building process in the DataMine Machine Learning Service is designed to be intuitive yet powerful, accommodating users with varying levels of expertise in data science and trading. Key steps include:

• Selection of Target Futures‍

Users start by choosing from 28 continuous futures, allowing them to focus their analysis on specific markets of interest.

• Predictor Selection‍

With access to over 47,152 data points from both CME and external sources, users can select predictors that are most relevant to their hypothesis, encompassing a wide range of categories like energy, economics, commodities, news, and more.

• Forecasting and Resource Allocation

‍The platform enables users to set forecasts for various days ahead, optimizing compute resources to ensure efficient and timely model training and prediction generation.

• Advanced Configuration

‍Through 43 advanced metrics, users can fine-tune their models, adjusting for factors that influence the accuracy and relevance of predictions, such as signal magnitude and forecast timeframes.

• Transparent and Iterative Development

‍The service promotes a transparent model development process, with detailed insights into driver analysis, model metrics, and individual feature importance. This supports iterative refinement, allowing users to continuously improve their models based on backtested performance and real-world outcomes.

Sophisticated Backtesting and Strategy Sandbox

One of the platform’s standout features is its sophisticated backtesting and strategy sandbox, allowing users to:

• Rewind models to past dates

Test how predictions would have performed historically.

• Generate auto strategies

Leverage ensembled predictions for more accurate forecasts.

• Assess strategy performance

Comprehensive metrics like ROI, Max Drawdown, and more, comparing against benchmarks for informed strategy adjustments.

Transparency and Customization

The DataMine ML Service emphasizes transparency in model metrics and outcomes, providing detailed breakdowns of individual and group feature importance. This ensures users understand the factors influencing model predictions, enabling refined strategy development. The service also offers pre-generated models via data feeds, governed by existing ILA, allowing for complete customization and privacy in strategy development.

Engagement and Commercial Opportunities

CME Group and QDT encourage potential users to engage with the platform, offering training and support to pilot customers. The service presents a unique white-label opportunity for firms, allowing private instances for customer bases and embedding in firm's websites or front ends. Additionally, it opens up new commercial implications and opportunities, from shared analytics to data feeds of complete model predictions.

‍