← Back to The IndexFinal ReviewGroup 11
Final Review

Review Nº 11

Predicting Data Centers Presence in Europe
AuthorsTeam JAM (ID 11)
27/30
Score
The submission covers the required workflow and offers useful feature interpretations, but model performance is moderate and several feature-engineering and validation choices need stronger justification.
Data Centres · Final Review

The Pros

1 Item
+
Explains data collection and NUTS3 harmonization; derives binary presence target and removes count to avoid leakage; uses stratified split; attempts thoughtful feature engineering for climate, per-capita, and correlated variables; compares Logistic Regression, Decision Tree, Random Forest, and SVM; selects a final model based on positive-class performance; discusses false negatives and limitations.

The Cons

1 Item
−
Feature engineering may remove important scale variables and introduce hard-to-interpret constructs; iterative imputation is not clearly constrained to training data; the final Random Forest AUC is only about 0.71 while Logistic Regression has higher AUC; model-selection rationale is not fully convincing; many interpretations rely on ambiguous climate mortality deltas; report needs clearer tables and stronger reproducibility.
↩ Back to The Index
Final Review · Group 11The IndexData Centres · 2026