Leveraging OCI’s AI and Machine Learning Services for Predictive Analytics

Setting Up Oracle AI Services

• Creating an AI Service Instance:
  • Log in to the OCI Console.
  • Navigate to AI Services → Create Service.
  • Select the service (e.g., Data Science, AI Platform) and follow the prompts to create an instance.
• Building a Machine Learning Model with OCI Data Science
  • Creating a Data Science Project:

oci data-science project create --compartment-id <compartment_OCID> --display-name "MyMLProject" --description "Project for predictive analytics"

Creating and Uploading Datasets:

oci data-science dataset create --compartment-id <compartment_OCID> --display-name "MyDataset" --data-location <object_storage_location> --format CSV

Creating a Model Training Job:

oci data-science job create --compartment-id <compartment_OCID> --project-id <project_OCID> --display-name "MyModelTrainingJob" --job-type "CUSTOM" --arguments '{"training_script":"<script_location>", "hyperparameters": {"learning_rate": 0.01}}'

Deploying and Using the Model

Deploying the Model:

oci data-science model-deployment create --compartment-id <compartment_OCID> --display-name "MyModelDeployment" --model-id <model_OCID> --deployment-config '{"instance_type": "VM.Standard2.2"}'

Invoking the Model Endpoint:

curl -X POST <model_endpoint_url> -H "Content-Type: application/json" -d '{"features": [value1, value2, ...]}'

Integrating Predictive Analytics into Business Workflows

• Creating Dashboards and Visualizations:
  • Use OCI Analytics Cloud or Oracle Analytics for visualization.
  • Example: Create a dashboard to visualize predictions and trends based on model output.

Automating Predictions:

• Set up automated workflows using OCI Functions to trigger model predictions based on new data.
• Example Function Deployment:
  
  

fn deploy --app myapp --image <docker_image> --env "MODEL_ENDPOINT_URL=<model_endpoint_url>"

Monitoring and Managing Models

• Monitoring Model Performance:
  • Use OCI Monitoring to track model performance metrics (e.g., accuracy, latency).
  • Example

oci monitoring metric-data list --compartment-id <compartment_OCID> --metric-name "model_accuracy"

Updating and Retraining Models:

• Periodically retrain the model with new data to improve performance.
• Example:

oci data-science job create --compartment-id <compartment_OCID> --project-id <project_OCID> --display-name "ModelRetrainingJob" --job-type "CUSTOM" --arguments '{"training_script":"<new_script_location>", "hyperparameters": {"learning_rate": 0.001}}'

Thank you
Osama