Applicazione del Machine Learning nella Valutazione del Solvency Ratio

Application of Machine Learning in the Evaluation of the Solvency Ratio Over the past decade, the insurance industry has undergone profound transformations, largely driven by the introduction of the Solvency II regulatory framework. This reform has standardized risk management and control processes, introducing key indicators such as the Solvency Ratio. The computation of this metric is a complex process that involves multiple corporate functions and requires an in-depth analysis of the main balance sheet components. Within this context, the increasing adoption of machine learning techniques is reshaping risk monitoring and management systems toward a more data-driven paradigm. This study aims to examine the application of statistical and machine learning methods to develop a predictive tool designed to identify insurance companies whose Solvency Ratios deviate significantly from expected values, thereby providing an early warning mechanism to support auditors, risk managers, and supervisory authorities. The analysis relies on a dataset comprising 50 non-life insurance companies, characterized by high dimensionality—many variables relative to the number of observations—which necessitates the use of advanced methodologies. The methodological framework unfolds in two stages: an initial phase focused on constructing aggregated indicators and defining prediction intervals, followed by a second phase in which more sophisticated machine learning models, including ensemble learning approaches, are applied to enhance the accuracy and robustness of the forecasts. The results show that the integration of advanced analytical techniques enables the development of effective tools for both internal company control and external supervision, contributing to the ongoing innovation within the insurance industry.