The recovery performance of a portfolio consisting of non-performing loans (NPLs) can be evaluated through a recovery curve, which illustrates the recovery rates over time, encapsulating both the recovery rate and the time taken for liquidation. In cases of significant portfolio heterogeneity, dividing the portfolio into multiple homogeneous groups, or clusters, and estimating a separate recovery curve for each, provides greater insight. This paper aims to determine the optimal portfolio partitioning, adopting an entropic fuzzy clustering approach, and to estimate smoothed recovery curves for each cluster using nonparametric statistical learning techniques.
Carleo, A., Rocci, R. (2025). Fuzzy clustering of NPL recovery curves. In IES 2025 - Innovation & Society: Statistics and Data Science for Evaluation and Quality (pp.72-78). Padova : CLEUP.
Fuzzy clustering of NPL recovery curves
Alessandra Carleo;
2025-01-01
Abstract
The recovery performance of a portfolio consisting of non-performing loans (NPLs) can be evaluated through a recovery curve, which illustrates the recovery rates over time, encapsulating both the recovery rate and the time taken for liquidation. In cases of significant portfolio heterogeneity, dividing the portfolio into multiple homogeneous groups, or clusters, and estimating a separate recovery curve for each, provides greater insight. This paper aims to determine the optimal portfolio partitioning, adopting an entropic fuzzy clustering approach, and to estimate smoothed recovery curves for each cluster using nonparametric statistical learning techniques.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
