Using Triggers and Actions in Workflows
Last updated July 29, 2024
Triggers and actions are fundamental components of automated workflows in Mostly AI, allowing you to control when and how your workflows execute. By strategically using triggers and actions, you can create more powerful and dynamic workflows that adapt to your specific needs and data patterns.
Triggers: Starting Your Workflow
Triggers initiate the execution of your workflow under specific conditions or events.
- Scheduled Triggers:
- Time-Based: Execute the workflow at regular intervals (daily, weekly, monthly, or specific times).
- Example: Run a model retraining workflow daily at 2:00 AM to update predictions based on the latest data.
- Event-Driven Triggers:
- Data Arrival: Trigger the workflow when new data arrives from a connected data source.
- Example: Start a workflow for processing and analyzing new customer information as soon as it's added to your database.
- Prediction Threshold: Initiate the workflow when predictions exceed a predefined threshold.
- Example: Trigger an alert workflow when a fraud detection model predicts a high probability of fraudulent activity.
- External Events: Integrate with external services or APIs to trigger workflows based on events outside of Mostly AI.
- Example: Start a workflow when a webhook notification is received from a third-party platform.
Actions: Performing Tasks Within Your Workflow
Actions are the steps performed within your workflow, driven by triggers.
- Data Manipulation:
- Data Cleansing: Remove duplicates, handle missing values, or transform data for model readiness.
- Feature Engineering: Create new features or modify existing ones to enhance model performance.
- Model Training:
- Retrain Models: Trigger model retraining with new data to update predictions.
- Hyperparameter Tuning: Optimize model hyperparameters for improved accuracy.
- Prediction Generation:
- Batch Predictions: Generate predictions for a large dataset at once.
- Real-Time Predictions: Generate predictions for individual data points as they arrive.
- Output Handling:
- Data Visualization: Create visual representations of prediction results.
- Report Generation: Generate reports based on prediction outcomes.
- Data Storage: Save predictions or intermediate results to files or databases.
- Notifications: Send alerts or notifications when specific conditions are met (e.g., exceeding a prediction threshold).
- External API Calls: Interact with other services and applications through API connections.
Real-World Application
- Automated Fraud Detection: A workflow could be triggered by an alert from the fraud detection model, initiating a sequence of actions that investigate the suspected fraud, report it to a security team, and block potentially fraudulent transactions.
By combining thoughtfully selected triggers and actions, you can build powerful workflows in Mostly AI that automate complex processes and enhance your machine learning operations.
Was this article helpful?