Standalone ML Model Training

Train ML models that make independent trading decisions

About Standalone ML Models

Standalone ML models make independent trading decisions (BUY/SELL/HOLD) without relying on rule-based models. They learn directly from historical price movements and outcomes.

Requirements: You need at least 500 recommendations with evaluated outcomes to train a model. The system will use all available historical data by default (up to 10 years).

More Data = Better Models: To maximize training data, expand your symbol watchlist to 2000+ symbols via the COLLECTION_SYMBOL_WATCHLIST environment variable. More symbols = more recommendations = better model performance. Check the ML Dashboard for current data statistics.

Train New Model

Select Model

Model Version

Use semantic versioning (e.g., 1.0.0, 1.0.1, 2.0.0)

Minimum Recommendations

Minimum data needed to proceed (default: 500)

Max Training Samples

Maximum samples to use (default: 50,000, accepts 1,000 to 100,000)

Use all available data (recommended)

When enabled, uses all historical recommendations regardless of date

Days Back (if not using all data)

How far back to look (max 10 years = 3650 days)

Error

Training Complete!

Model: v

Test Accuracy: %

Test F1 Score: %

Redirecting to model registry in 3 seconds...

Trained Standalone Models

Loading models...

No standalone ML models trained yet

Train your first model above to get started