PENSAMENTO COMPUTACIONAL E APRENDIZADO DE MÁQUINA: DESAFIOS NA PREDIÇÃO DO DESEMPENHO ESCOLAR COM REDES NEURAIS ARTIFICIAIS

• DOI: https://doi.org/10.22533/at.ed.325112612012

• Palavras-chave: Redes Neurais Artificiais, Aprendizado Profundo, Inteligência Artificial, Treinamento de Modelos, Aplicações Multidisciplinares

• Keywords: Artificial Neural Networks, Deep Learning, Artificial Intelligence, Model Training, Multidisciplinary Applications

• Abstract: This article presents a systematized, clear, and technically grounded synthesis of the main concepts that structure the field of Artificial Neural Networks (ANNs). The fundamentals of the artificial neuron and network architectures are progressively addressed, as well as essential aspects of the modeling workflow, including data preprocessing and normalization, hyperparameter definition, activation functions, and the stages of training, validation, and testing. The text also discusses critical phenomena affecting model performance, such as overfitting and underfitting, highlighting their causes, implications, and mitigation strategies. Beyond the theoretical foundations, multidisciplinary applications of ANNs are presented in areas such as healthcare, education, industry, agriculture, finance, and emerging technologies, emphasizing their central role in contemporary intelligent systems. The objective of this work is to provide an integrated, didactic, and accessible perspective to support students and professionals in understanding the principles of deep learning and its practical implications in the development of artificial intelligence–based solutions. This work concludes with a final discussion and directions for future research.