Mapeamento sistemático sobre técnicas de monitoramento de faltas baseadas em mineração de dados

Abstract

The increasing complexity and interconnection of modern software systems makes fault detection a significant challenge, as traditional testing techniques have proven insufficient, particularly in dynamic and large-scale systems. To address this issue, various real-time monitoring approaches based on data mining have been proposed. These approaches, through continuous system analysis and the use of anomaly detection techniques, are capable of identifying faults in software systems. This study presents a systematic mapping of the literature on these monitoring techniques, aiming to identify implementation challenges and provide a comprehensive overview of the current state of the research. The mapping was conducted using the Scopus database, and 20 studies were selected to help answer the research questions defined in this work. The analysis enabled the identification of not only the main detection techniques but also the types of monitored systems, fault categories, and evaluation metrics used in the proposed approaches. Among the main challenges identified are the difficulty in labeling data for training supervised models, the complexity of interpreting the generated models, and the need for computationally efficient monitoring.