Eficiência técnica em assentamentos rurais do Ceará: análise envoltória de dados, clusters e aprendizado de máquina

Abstract

The objective of this thesis is to analyze the technical efficiency of family-run production units in rural settlements in the state of Ceará, seeking to identify its determinants and propose recommendations for the sustainability of these territories. To this end, a quantitative methodological approach was employed, relating Data Envelopment Analysis (DEA), cluster analysis (K-means), and machine learning algorithms (Random Forest, SVM, and XGBoost). The research used secondary data from 923 plots present in 16 Settlement Projects (PAs), distributed across ten municipalities. This data was also explored qualitatively, aiming at a socioeconomic and productive characterization of the settlements that comprise the sample. The DEA results revealed notable heterogeneity in performance, with 60% of the municipalities operating on the efficiency frontier and 40% presenting inefficiencies of up to 26.2%. The cluster analysis segmented the municipalities into three distinct profiles (small, medium, and large), validating the diversity of productive contexts. Machine learning models demonstrated high predictive capacity, with Random Forest achieving 89.7% accuracy in efficiency classification and XGBoost explaining 80.7% of the variation in efficiency score (R²). The area cultivated with corn emerged as the most important predictor, followed by access to water and diversification for livestock (poultry, cattle, and swine), which proved to be critical factors for productive success. It is concluded that the sustainability of settlements is intrinsically linked to increased technical efficiency, which, in turn, depends on overcoming socioeconomic deficits, such as the lack of water infrastructure. The mere possession of land proves insufficient without complementary investments. The research findings provide support for the formulation of customized public policies, with specific recommendations for each municipality profile, aiming to promote fairer, more resilient, and productive rural development in the semi-arid region of Ceará.