← Back to The IndexFinal ReviewGroup 2
Final Review

Review Nº 02

Predicting Data Center Presence in European NUTS3 Regions: A Machine Learning Approach
AuthorsTeam 2 / GENYZ
27/30
Score
Good end-to-end pipeline with multiple models and useful interpretation, but the model-selection logic, PCA use, and some methodological explanations are inconsistent or under-justified.
Data Centres · Final Review

The Pros

>
+
Clear motivation and related work; documents data sources and preprocessing steps; uses train/test split before scaling; compares several classifiers; tunes hyperparameters; reports multiple metrics; includes permutation importance and partial dependence; discusses limitations and links results to hypotheses.

The Cons

−
Logistic regression appears to have the highest AUC in the table, yet SVM-RBF is selected without a fully convincing rationale; PCA-transformed features reduce interpretability for the main project goal; threshold tuning is confusingly described as “0.5”; some feature interpretations, especially climate losses and U-shaped electricity price effects, are speculative.
↩ Back to The Index
Final Review · Group 2The IndexData Centres · 2026