Beyond National Accuracy: A Neutrosophic Geographic Evidence Audit of Preschool-Attendance Prediction in Ecuador (ENSANUT 2018)

Authors

• Cindy Patricia Campoverde Vera, Maestria en Analitica de Datos, Universidad Politecnica Salesiana, Ecuador

Abstract

Predictive models built on national surveys are usually summarized by one aggregate performance figure, yet a national metric may conceal large differences in how the model behaves across territories. We examined this problem with the 2018 Ecuadorian National Health and Nutrition Survey (ENSANUT 2018). A conventional Random Forest trained on five child and household variables reached a national accuracy of 64.83% against a majority-class baseline of 54.40%. This aggregate concealed substantial geographic heterogeneity: provincial accuracy ranged from 56.1% (Bolivar) to 78.0% (Canar). To represent this evidence without collapsing it into one number we formalize a Neutrosophic Geographic Evidence Audit (NGEA), inspired by the evidence-separation principle of Neutrosophic Paraconsistent Logic. The audit changed the qualitative verdict of 9 of 24 provinces relative to a single-split accuracy with confidence interval. The study is observational and cross-sectional; heterogeneity of model performance is not interpreted as territorial inequality.

Recommended Citation

Campoverde Vera, Cindy Patricia. "Beyond National Accuracy: A Neutrosophic Geographic Evidence Audit of Preschool-Attendance Prediction in Ecuador (ENSANUT 2018)." Neutrosophic Sets and Systems 101, 1 (2026). https://digitalrepository.unm.edu/nss_journal/vol101/iss1/4