AI Prediction Node in Pathways

The AI Prediction node in FojiSoft Pathways is used to generate predictions based on a preconfigured AI model. This node allows workflows to integrate machine learning-based predictions seamlessly, supporting decision-making and automation in complex processes.

---

Node Properties

1. Name

• Description: A unique identifier for the AI Prediction node.
• Usage: Provide a clear and descriptive name that reflects the purpose of the prediction (e.g., "Claim Approval Prediction" or "Patient Risk Assessment").

---

2. Configuration

• Description: Defines the AI model and its associated settings to be used for making predictions.
• Components:
  1. Model Selection:
     • Choose from preconfigured AI models available in Forge AI.
     • Models may include classification, regression, clustering, or custom-trained models.
  2. Input Mapping:
     • Map workflow variables or data inputs to the features required by the AI model.
  3. Prediction Type:
     • Specify the type of prediction the AI model will perform (e.g., binary classification, multi-class classification, regression).

---

3. Output Variable

• Description: The variable name where the prediction results will be stored.
• Usage: Define a unique variable name to store the output, such as claim_approval_status or predicted_risk_score.
• Example:
  • If the model predicts whether a claim will be approved:
    • Output Variable: claim_approval
    • Prediction Output: Approved or Rejected

---

How It Works

1. Input Configuration:
   • Define the input features required by the AI model.
   • Ensure inputs are properly mapped to relevant workflow data (e.g., patient details, insurance policy, historical records).
2. Model Prediction:
   • The selected model processes the inputs and generates a prediction.
   • Predictions can include probabilities (e.g., likelihood of approval) or classifications (e.g., "Approved" vs. "Rejected").
3. Output Handling:
   • Store the prediction results in the specified Output Variable.
   • Use the results in downstream nodes for further actions, such as notifications or data updates.

---

Example Usage

Scenario: Insurance Claim Approval Prediction

1. Name: Claim Approval Prediction
2. Model Configuration:
   • Selected Model: "ClaimApprovalModel"
   • Input Features:
     • claim_amount
     • policy_type
     • patient_age
   • Prediction Type: Binary Classification (Approved or Rejected)
3. Output Variable: claim_status

Workflow Process:

1. Input claim details into the workflow.

The AI model predicts the likelihood of approval:

{

 "claim_status": "Approved",

 "confidence": 0.92

}

1. Output variable claim_status is used in the next steps of the workflow to notify staff or proceed with claim submission.

---

Best Practices

1. Select the Right Model:
   • Use models that are pre-trained for your specific task, or train a custom model using Forge AI.
2. Map Inputs Correctly:
   • Ensure input variables align with the model’s requirements for optimal performance.
3. Define Clear Output Variables:
   • Use meaningful variable names to make workflows easy to debug and maintain.
4. Test the Workflow:
   • Run test predictions to validate the model's outputs before deploying the workflow.