← Back to The IndexFinal ReviewGroup 7
Final Review

Review Nº 07

Drivers of Data Center Location in European NUTS3 Regions: An Integrated Econometric and Machine Learning Pipeline
AuthorsGroup 7
29/30
Score
Strong integrated pipeline with good model variety and performance, but threshold selection on the test set and SMOTE usage need correction to avoid optimistic evaluation.
Data Centres · Final Review

The Pros

1 Item
+
Clear dataset and target definition; good literature-based hypotheses; uses transformations, median imputation, correlation diagnostics, logistic regression, Random Forest, Gradient Boosting, and regularized logistic regression; reports both threshold metrics and ROC-AUC; recognizes water exploitation as a constraint; performance is competitive, especially Gradient Boosting F1 and RF AUC.

The Cons

1 Item
−
Selecting the Gradient Boosting threshold using the test set is a major evaluation flaw; SMOTE is arguably unnecessary with a 44/56 class split and may distort regional feature distributions; the report sometimes uses “I” instead of team/report voice; source “PCH” likely should be PCH/IXP directory clarified; spatial dependence and country-level confounding are not sufficiently tested.
↩ Back to The Index
Final Review · Group 7The IndexData Centres · 2026