Making Predictions Guide¶
This guide covers how to use your trained Arc models to make predictions on new data.
Overview¶
Once you've trained and evaluated a model, you can use it for inference (making predictions). Arc makes this simple: 1. Load new data - Get data you want predictions for 2. Run prediction - Apply your trained model 3. View results - Access predictions via SQL
Quick Start¶
Make predictions on new data:
Or with the /ml predict command:
Using the /ml predict Command¶
Basic Prediction¶
Predictions are saved to <model_name>_predictions table by default.
Custom Output Table¶
Prediction Workflow¶
Prepare Input Data: Ensure new data has the same features as training data with identical preprocessing.
Run Prediction: /ml predict --model my_model --data new_data
View Predictions: /sql SELECT * FROM my_model_predictions LIMIT 10
Understanding Prediction Outputs¶
Classification: Returns predictions with probabilities and predicted classes.
Regression: Returns predicted values.
Example: Batch Scoring¶
Query results with /sql SELECT * FROM churn_scores WHERE churn_probability > 0.7
Next Steps¶
- Model Evaluation Guide - Evaluate prediction quality
- Model Training Guide - Retrain with new data
- API Reference - All prediction commands
Related Documentation¶
- Arc-Graph Specification - Understand model architecture
- Data Loading Guide - Load new data for prediction