Data analytics-based approaches for spatial decision support in energy infrastructure and criminology

Abstract

Like other countries in the Global South, Brazil faces concurrent strategic challenges: historically high and uneven levels of violence and the need to accelerate the energy transition. In tourist cities and in the Northeastern interior, the human and social costs of violence persist; at the same time, expanding renewable energy requires siting decisions that reconcile resource potential, infrastructure, constraints, and impacts. In this context, this dissertation advances an applied agenda in two domains: (i) documenting and explaining the spatiotemporal organization of homicides across distinct territorial arrangements; and (ii) proposing a decision-support framework for evaluating wind project locations. The central objective is to transform data into replicable, decision-oriented evidence, producing interpretable outputs that support time- and place-sensitive actions. In the field of criminology, research investigates a tourist capital with severe violence and the rural–urban continuum of a Northeastern state, examining temporal patterns, persistent geographic concentrations, and their relationships with contextual factors. Results indicate recognizable routine patterns, stable clusters in specific sectors, and robust associations between structural variables and homicide incidence, providing interpretation beyond descriptive mapping. At the state scale, the analysis reveals specific patterns of lethal violence, challenging urban-centric theories by documenting elevated rates and situational dynamics in rural and remote territories. In the energy transition domain, it presents a decision support framework based on preference learning and produces decision rules to sort areas, integrating multiple decision-makers’ preferences and reducing cognitive effort. A case study in a Northeastern Brazilian state is presented and shows reliable and explainable results, with improved coverage and discrimination as rules are refined, signaling adaptability to other territories and renewable sources. Overall, the dissertation integrates spatiotemporal diagnostics, explanatory modeling, and decision support to produce applicable and replicable knowledge. Contributions lie in explaining temporal and geographic patterns of homicide in the Global South—generating knowledge about urban and rural environments—and in structuring wind power project evaluation processes based on preference learning, yielding meaningful results. The proposed procedures are adaptable to different contexts, consolidating a practical basis for informed decision-making in public safety and energy planning.