DEFESA DE DISSERTAÇÃO DE MESTRADO Nº 345

Aluno: Pedro Natanael Firmino da Silva

Título: “Alliance Dynamics and Vote Outcome Prediction in the Brazilian Chamber of Deputies"

Orientador: Dimas Cassimiro do Nascimento Filho

Coorientador: Bruno Costa e Silva Nogueira (UFAL)

Examinador Externo: Tiago Figueiredo Vieira (UFAL)

Examinador Interno: Diego Marconi Pinheiro Ferreira Silva

Data-hora: 25 de março de 2026 às 15h

Local: Formato remoto - Google meet

---

Resumo:

         "Legislative roll-call records provide a high-resolution view of how representatives align on institutionally consequential decisions, supporting transparency, accountability, and evidence-based monitoring. In multiparty settings such as the Brazilian Chamber of Deputies, however, coalition structure can be blurred by weakly polarized sessions and noisy similarity edges, and outcome forecasting can be overestimated when temporal leakage is not carefully avoided. This dissertation presents a computational framework that integrates two complementary goals. In Chapter 5, we construct yearly voting-similarity networks among deputies and apply community detection to identify cohesive blocs interpretable as alliances. We propose a noise-reduction strategy based on polarization-aware session filtering and weak-edge pruning before applying the Leiden algorithm, enabling clearer communities and longitudinal analysis of bloc composition from 2004–2023. In Chapter 6, we formulate proposition-level outcome prediction (approved vs. rejected) as a realistic forecasting task under strict chronological evaluation. The VOTE-RAP model combines institutional signals (government orientation), leakage-safe temporal approval patterns (party popularity and historical approval rate), and authorship-based coalition-size proxies, delivering interpretable performance with emphasis on identifying rejections. The empirical results show that explicit noise handling improves modularity (10.95% over a backbone baseline) and that leakage-safe temporal signals enable accurate outcome forecasting (AUROC 0.908; rejected-class F1 0.703) under chronological evaluation. The proposed methods can benefit journalists, oversight institutions, legislative observatories, and public-sector analysts who need transparent tools to track coalition shifts and prioritize propositions for monitoring. "