Assessment Checklist

Before you submit your M9 final assessment, check every box below to ensure you have covered the full ML workflow.

The MVP Submission

• Have I formulated a clear ML business problem?
• Did I acquire, clean, and engineer features appropriately?
• Have I trained at least one baseline and one advanced model?
• Is there rigorous validation (CV, training vs test analysis)?
• Have I explicitly translated the model metric (Accuracy/RMSE) into a business metric (£ saved or time reduced)?
• Did I communicate the results effectively with a visualisation?

Data Preparation

• Missing values handled (dropped, imputed, or flagged)
• Categorical features encoded (One-Hot, Ordinal, or Target Encoding)
• Numerical features scaled where required (StandardScaler for distance-based models)
• No data leakage — preprocessing fitted on training data only

Modelling

• At least two algorithms compared (e.g., Logistic Regression vs Random Forest)
• Hyperparameters tuned systematically (GridSearch, RandomSearch, or Optuna)
• Cross-validation used — not a single train/test split
• Overfitting checked (training vs test score gap)

Evaluation

• Appropriate metric chosen (not just accuracy — consider F1, ROC AUC, RMSE)
• Confusion matrix or classification report included (for classification)
• Residual analysis included (for regression)
• Confidence intervals or standard deviation reported

Communication

• Feature importance visualised (SHAP, permutation, or tree-based)
• Results presented in plain English for a non-technical audience
• Business impact quantified (ROI, cost savings, efficiency gains)
• Limitations and next steps acknowledged

KSB Mapping

KSB	Description	How This Addresses It
K1	Context of Data Science	Understanding where ML sits within the broader discipline
S3	Programming languages and tools	Setting up the development environment and dependencies
B6	Commitment to keeping up to date	Engaging with current ML resources and research