Type:	Monografia
Title:	Tree-based learning for game outcome prediction
Author:	Bonelli, Djonatan Riquelme Clein
First advisor:	Grando, Felipe
Resume:	Predicting outcomes in complex multi-agent games is challenging due to imper- fect information, stochastic events, and strategic interactions. We investigate interpretable tree-based models for outcome prediction in Citadels, a strategic board game that serves as a testbed for multi-agent dynamics. Using Optuna for hyperparameter optimization, we configure Decision Trees and Random Forests, with the latter consistently outperforming single trees. Prediction accu- racy exceeds 60% in early game rounds and surpasses 90% in later rounds, while metric-specific optimization highlights trade-offs among precision, recall, and F1 score. showing that combinations of a few features can yield strong predictive signals when interaction effects are considered. Our results demonstrate that interpretable tree-based models can combine robust predictive performance with transparent explanations, offering insights into strategic behavior and informing the broader study of decision-making in complex multi-agent environments.
Keywords:	Inteligência artificial Predição Jogos de tabuleiro Sistemas multiagentes Algoritmos
Language:	por
Country:	Brasil
Publisher:	Universidade Federal da Fronteira Sul
Acronym of the institution:	UFFS
College, Institute or Department:	Campus Chapecó
Type of Access:	Acesso Aberto
URI:	https://rd.uffs.edu.br/handle/prefix/9176
Issue Date:	2025
Appears in Collections:	Ciência da Computação

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.