Labcity UFPA | Inteligência Artificial & Cidades Inteligentes

Resumo

The accelerated growth of urbanization imposes increasing challenges on public management, demanding innovative solutions to optimize urban services and strengthen the bond between citizens and municipal authorities. This article presents a Smart City Architecture (SCA) developed for the municipality of Canaã dos Carajás in the Brazilian Amazon, focusing on the application of state-of-the-art Artificial Intelligence (AI) techniques to minimize the communication gap between citizens and municipal authorities. The proposed framework integrates a mobile application for incident reporting, a web system for administrative management, and machine learning models utilizing BGE-M3 semantic embeddings to automatically classify urban service requests in Brazilian Portuguese. To ensure algorithmic fairness and mitigate class imbalance, the Synthetic Minority Over-sampling Technique (SMOTE) was strategically applied. Five classification algorithms were evaluated: XGBoost, Logistic Regression, Random Forest, Support Vector Machine (SVM), and Bidirectional Long Short-Term Memory (BiLSTM). The architecture was designed under a privacy-by-design paradigm, ensuring local data sovereignty and compliance with the Brazilian General Data Protection Law (LGPD) and the European General Data Protection Regulation (GDPR) through an on-premise processing approach. A case study conducted in Canaã dos Carajás (Brazil) demonstrated the efficiency of the approach, revealing high performance in categorizing incidents (with AUC up to 0.98 and F1-scores reaching 0.94) and highlighting the potential of AI to support public decision-making while significantly improving the recall of underrepresented municipal departments. The results indicate that AI-based solutions, when combined with semantic richness and ethical governance, can foster more efficient, participatory, and data-driven management, bringing citizens and municipal administration closer.