Desenvolvimento de um modelo híbrido de inteligência artificial como ferramenta de prevenção ao risco de insolvência empresarial

Abstract

Given the growing academic interest in the topic of "Bankruptcy Prediction" and recognizing its socioeconomic relevance, this study aims to develop a hybrid Artificial Intelligence (AI) model as a tool for preventing business insolvency risks. It is worth noting that the proposed model comprises two distinct techniques that operate complementarily, specifically an Artificial Neural Network (ANN) and an Expert System (ES). The ANN is primarily designed to assess the insolvency risk of business assets, while the ES focuses on analyzing potential financial anomalies and subsequently recommending managerial interventions to preserve the economic integrity of the business model. The proposal is the result of an extensive literature review, which highlighted the need for a new methodological approach, given that the field is saturated with discussions that fail to go beyond the bias of "predictive efficiency." It is important to emphasize that the dataset used for training and testing the mathematical models was sourced from Brazilian companies listed on the Ibovespa index. Additionally, the development of the ES benefited from contributions by three financial professionals. Finally, the results obtained from applying the intelligent models demonstrate the feasibility of using hybrid techniques to build complex systems that not only identify financial risks at an early stage but also recommend corrective actions. In summary, the predictive system achieved 95% accuracy in identifying insolvency risks up to two years in advance, enabling timely diagnoses and actionable inferences by the ES. Therefore, it is believed that the model will expand the perspective of practical applications in the business context while fostering the emergence of a new generation of studies related to bankruptcy prediction. [resumo fornecido pelo autor]