This guide will walk you through the features and capabilities of the Machine Learning Toolkit, covering how to set up models, optimize them, and integrate them into your IoT and AI workflows.
Rayven.io’s Machine Learning Toolkit enables you to build, deploy, and optimize machine learning models directly within your data workflows. Whether you're using IoT devices or historical data, the toolkit supports real-time decision-making, anomaly detection, forecasting, and more.
This guide covers how to:
-
Access the toolkit
-
Configure different types of models
-
Optimize them in real-time
-
Integrate them into workflows and dashboards
1. Accessing the Machine Learning Workbench
To begin:
-
Navigate to AI & Machine Learning in the Rayven platform.
-
Click Machine Learning Workbench to open the configuration environment.
2. Model Types Available
Rayven offers three model creation options:
a. Ready-to-Go Models
Pre-built models for common use cases like:
-
Anomaly detection
-
Predictive maintenance
-
Forecasting
How to use:
-
Select Ready-to-Go Models.
-
Choose a model type.
-
Connect to live or historical data inputs.
-
Deploy and begin receiving predictions.
b. Custom Model Creation
Create models tailored to your specific data and objectives.
Steps:
-
Choose Custom Model Creation.
-
Define the algorithm (e.g., regression, classification).
-
Input training data from your tables or devices.
-
Set expected outputs (e.g., score, label).
-
Save and deploy.
c. Python Model Import
Import a pre-trained Python model (e.g., TensorFlow, Scikit-learn).
Steps:
-
Select Python Model Import.
-
Upload the model file.
-
Map input/output fields to Rayven’s data sources.
-
Connect to live streams or batch inputs.
3. Real-Time Optimization
Enable models to self-improve as new data arrives.
To enable:
-
Toggle Real-Time Optimization in the model settings.
-
Link live data streams (e.g., sensors).
-
Ensure the model retrains automatically or updates on-the-fly.
This is critical for applications like predictive maintenance and real-time anomaly detection.
4. Seamless IoT Integration
Rayven’s platform allows direct integration between IoT devices and machine learning models.
How to set up:
-
In the model’s Data Input section, select connected IoT devices or sensor streams.
-
Map relevant fields (e.g., temperature, vibration, current).
-
Use model outputs for real-time alerts or dashboard displays.
5. Workflow Integration
Machine learning models can be embedded in your Rayven workflows to automate decisions and actions.
Steps:
-
Open the Workflow Editor.
-
Add a node and select Machine Learning Model.
-
Choose the model you configured.
-
Define triggers or conditions (e.g., “If predicted score > 0.8, send email”).
-
Save the workflow.
6. Monitoring Model Performance
Tools are provided to ensure your model stays accurate and effective:
-
Accuracy Tracking: View performance in real-time.
-
Retraining: Easily retrain with newer datasets.
-
Dashboards: Visualize predictions, accuracy metrics, and KPIs.
7. Common Use Cases
Use Case | Description |
---|---|
Predictive Maintenance | Detect potential equipment failure early using sensor data. |
Anomaly Detection | Flag outliers and abnormal patterns that may indicate problems or threats. |
Forecasting | Predict demand, energy use, or other metrics using time-series data. |
Q&A
Q: Can I use models built outside Rayven?
Yes, via Python Model Import. Upload your .pkl
or other supported model and map its inputs/outputs.
Q: Do models support real-time retraining?
Yes. Toggle Real-Time Optimization and link live data streams.
Q: Where can I see model results?
Results can be displayed in dashboards or used to trigger actions in workflows.
Q: What formats are supported for training data?
Any Rayven-connected tables or IoT data streams can serve as training inputs.
By integrating machine learning into your Rayven.io workflows, you unlock real-time intelligence and automation. Whether you're deploying a pre-built model or training a custom one, the Machine Learning Toolkit provides all the capabilities you need—from setup to monitoring—to transform your data into predictive power.