MACHINE LEARNING APPROACHES TO DEFAULT PREDICTION IN THE AUTOMOTIVE SECTOR

Abstract

The objective of this research was to conduct a bibliometric analysis of the international literature on the use of machine learning techniques for default prediction, with a focus on the automotive sector. A total of 1.026 documents published between 2010 and 2025 in the Scopus database were examined, aiming to identify thematic clusters, emerging trends, and methodological challenges. The results indicate a rapidly consolidating scientific field, characterized by international expansion, thematic diversification, and methodological maturity. The evidence is organized into four main axes: (i) the consolidation of machine learning as the central approach in credit risk modeling, gradually replacing traditional statistical methods; (ii) methodological expansion, with emphasis on neural networks, reinforcement learning, and ensemble models; (iii) growing appreciation for algorithmic interpretability and transparency; and (iv) integration of multiple data sources, combining financial, textual, and technological information. Moreover, the study identified a lack of systematic investigations into the application of machine learning models to default prediction specifically within the automotive sector, revealing a relevant gap in the literature. The implications of these findings include the construction of a structured overview of key trends, authors, and methodologies, providing theoretical and empirical insights to advance future research and improve predictive credit risk models.