2026-10-06T13:38:47Z https://api.figshare.com/v2/oai

oai:figshare.com:article/34029324 2026-09-30T04:00:20Z category_26383 category_26353 category_27634 category_25978 category_29173 category_29890 item_type_3 month_year_09_2026

Sobrecostos en obras públicas de infraestructura del Perú: base de obras, variables predictoras y resultados, 2018–2023 Miluska Rodriguez-Saavedra (23594950) Civil engineering not elsewhere classified Construction engineering Public administration International economics Machine learning not elsewhere classified Statistical data science public infrastructures Cost Overruns public works contractors construction management. public investment bodies Machine learning accelerators <p dir="ltr">This dataset contains project-level data on public infrastructure works in Peru, prepared for the prediction and analysis of cost overruns using machine-learning methods. The analytical sample covers public works with approved technical files between 2018 and 2023.</p><p dir="ltr">The database includes information on project identification, sector, geographic location, level of government, viable investment cost, approved technical-file cost, project characteristics, institutional variables, historical execution indicators, and variables derived for the analysis of cost deviations.</p><p dir="ltr">The main analytical outcome is the percentage cost overrun between the approved technical-file cost and the viable investment cost. The dataset includes the original cost-overrun measure, cleaned and winsorized versions, a log-transformed outcome, and binary indicators identifying critical cost overruns at alternative thresholds.</p><p dir="ltr">The database also contains data-quality indicators and variables prepared to reduce the risk of information leakage in predictive modelling. Variables that may not have been observable at the intended prediction point are retained for robustness analyses rather than for the main predictive specification.</p><p dir="ltr">A secondary analytical subset contains public works with sufficient information for delay analysis. Missing values were not artificially imputed, and the original variables were preserved alongside cleaned or transformed versions.</p><p dir="ltr">The dataset was constructed from official public investment and public works information from Peru and is intended to support reproducible research using XGBoost, SHAP, regression, classification, and related predictive methods for the study of cost overruns in public infrastructure.</p> 2026-09-30T04:00:20Z Dataset Dataset 10.6084/m9.figshare.34029324.v1 https://figshare.com/articles/dataset/Sobrecostos_en_obras_p_blicas_de_infraestructura_del_Per_base_de_obras_variables_predictoras_y_resultados_2018_2023/34029324 CC BY 4.0