Oracle Cloud Infrastructure Data Science Professional — Lesson 3: Model Selection, AutoML, Evaluation, Explainability, and Model Catalog

Learning Objectives

• Explain the purpose and operating context of algorithm selection, training strategies, hyperparameters, and distributed awareness.
• Apply AutoML, tuning, model comparison, and experiment ranking and metrics, imbalance, thresholds, uncertainty, and error analysis in a controlled environment.
• Validate outcomes and troubleshoot risks involving explainability, fairness, model catalog metadata, approvals, and risk.

Detailed Topics

• algorithm selection, training strategies, hyperparameters, and distributed awareness
• AutoML, tuning, model comparison, and experiment ranking
• metrics, imbalance, thresholds, uncertainty, and error analysis
• explainability, fairness, model catalog metadata, approvals, and risk

Instructor Demonstration

Demonstrate algorithm selection, training strategies, hyperparameters, and distributed awareness for an enterprise model portfolio designed for secure collaboration, repeatable delivery, and monitored inference; capture the exact configuration, code, command, transaction, query, dashboard, model, schedule, or report evidence that proves the expected result.

Hands-On Exercise

Complete a controlled Oracle Cloud Infrastructure Data Science Professional exercise covering AutoML, tuning, model comparison, and experiment ranking and metrics, imbalance, thresholds, uncertainty, and error analysis; introduce an exception involving explainability, fairness, model catalog metadata, approvals, and risk, diagnose it from evidence, apply the smallest reversible correction, and rerun acceptance checks.

1. Confirm scope, prerequisites, identities, dependencies, data, risks, and expected outcome.
2. Perform the course-specific configuration, development, administration, analysis, or business-process workflow.
3. Capture commands, settings, records, logs, screenshots, reports, metrics, or test results.
4. Compare the observed result with acceptance criteria and document corrective or preventive action.

Scenario Analysis

• Identify the technical or business decision and the stakeholders affected by it.
• Compare security, performance, reliability, compliance, cost, usability, and support trade-offs.
• Define rollback, recovery, communication, evidence, ownership, and escalation requirements.

Troubleshooting Guidance

Diagnose issues involving algorithm selection, training strategies, hyperparameters, and distributed awareness, AutoML, tuning, model comparison, and experiment ranking, metrics, imbalance, thresholds, uncertainty, and error analysis, explainability, fairness, model catalog metadata, approvals, and risk without making an unverified production change. Separate symptoms from causes, gather time-correlated evidence, test a reversible hypothesis, validate the correction, and record recurrence controls.

Validation Checklist

• The requirement, environment, inputs, dependencies, and expected result are documented.
• The result is reproducible and supported by product-specific evidence.
• Normal, boundary, failure, security, performance, and operational conditions are reviewed.
• Exceptions, rollback, ownership, support actions, and the next improvement priority are recorded.

Lesson Outcome

Learners can apply Model Selection, AutoML, Evaluation, Explainability, and Model Catalog in a realistic Oracle Cloud Infrastructure Data Science Professional scenario and defend the result with documented evidence.