Oracle Cloud Infrastructure Data Science Professional (2025): Hands-on Workshop — Lesson 2: Data profiling, preparation, feature engineering, imbalance treatment, experiment design, training, hyperparameter search, and model evaluation

Learning Objectives

• Identify the configuration and data choices that shape Data profiling, preparation, feature engineering, imbalance treatment, experiment design, training, hyperparameter search, and model evaluation.
• Run normal and boundary tests and record the outcome in feature pipeline and experiment comparison report.
• Separate root cause from symptom when investigating cross-validation results overstating performance because data leakage was introduced during preparation.

Detailed Topics

• Data profiling
• Preparation
• Feature engineering
• Imbalance treatment
• Experiment design
• Training
• Hyperparameter search
• And model evaluation
• Decision ownership, prerequisites, and acceptance evidence for Data profiling, preparation, feature engineering, imbalance treatment, experiment design, training, hyperparameter search, and model evaluation
• Relevant access, data, integration, performance, recovery, and support considerations

Instructor Demonstration

The instructor configures and tests Data profiling, preparation, feature engineering, imbalance treatment, experiment design, training, hyperparameter search, and model evaluation within a scenario aimed at a reproducible OCI Data Science workflow that prepares data, trains and deploys models, automates lifecycle tasks, and monitors business-facing predictions. The walkthrough contrasts a valid setup with a flawed one and uses feature pipeline and experiment comparison report to make the difference explicit.

Hands-On Exercise

Configure and test Data profiling, preparation, feature engineering, imbalance treatment, experiment design, training, hyperparameter search, and model evaluation with representative records. Use “cross-validation results overstating performance because data leakage was introduced during preparation” as the troubleshooting challenge, compare competing hypotheses, retest only the affected path, and complete feature pipeline and experiment comparison report for peer review.

1. Confirm scope, access, sample records, and acceptance criteria.
2. Perform the configuration or transaction using a controlled sequence.
3. Inspect outputs and downstream effects rather than relying on screen messages alone.
4. Close with a support note, preventive action, and reproducible validation step.

Scenario Analysis

• Clarify the business trigger, affected roles, required records, and acceptable result for Data profiling, preparation, feature engineering, imbalance treatment, experiment design, training, hyperparameter search, and model evaluation.
• Compare at least two implementation or correction choices and identify which evidence would distinguish them.
• Decide how feature pipeline and experiment comparison report should be reviewed, approved, retained, and handed to the next owner.

Troubleshooting Guidance

When results from Data profiling, preparation, feature engineering, imbalance treatment, experiment design, training, hyperparameter search, and model evaluation are unexpected, first reproduce the issue with the same role, record, date, and transaction context. Inspect feature pipeline and experiment comparison report, then test whether “cross-validation results overstating performance because data leakage was introduced during preparation” explains the evidence before changing configuration. Retest the corrected path and one unaffected path to guard against regression.

Validation Checklist

• The scope of Data profiling, preparation, feature engineering, imbalance treatment, experiment design, training, hyperparameter search, and model evaluation is clear, including what is deliberately excluded from the test.
• Evidence in feature pipeline and experiment comparison report supports each material decision and outcome.
• A negative test demonstrates that controls reject or flag an invalid condition.
• Open risks, owners, due actions, and production safeguards are documented.

Lesson Outcome

Learners can apply Data profiling, preparation, feature engineering, imbalance treatment, experiment design, training, hyperparameter search, and model evaluation, explain the decisions made, and defend the result with feature pipeline and experiment comparison report.