← Back to The IndexFinal ReviewGroup 9
Final Review

Review Nº 09

Regional Determinants of Data Center Location in Europe At NUTS3 level
AuthorsGroup 9 - The Oracle
28/30
Score
Very well-written and theoretically grounded submission with strong interpretation, but feature selection using the full dataset creates leakage risk and weakens the evaluation claims.
Data Centres · Final Review

The Pros

1 Item
+
Excellent abstract and motivation; clear hypotheses; good dataset description; strong discussion connecting results to economic scale, connectivity, and environmental variables; compares logistic regression and Random Forest; reports held-out and CV AUC; interprets counterintuitive population and pollution effects thoughtfully.

The Cons

1 Item
−
Feature selection is performed using Pearson correlation with the target on the full dataset before splitting, which leaks test information; environmental conclusions are strong despite variables being proxies; no spatial/country validation; excluded log_ixp_count due to collinearity with has_ixp needs a clearer modelling justification; no resampling/class-weighting is acceptable but should be justified by target metric choices.
↩ Back to The Index
Final Review · Group 9The IndexData Centres · 2026