Selected work

Representative Papers

Surrogate Graph Partitioning for Spatial Prediction

Oct 09, 2025

This paper addresses the poor interpretability of black-box models in spatial prediction. We propose a graph-partitioning-based spatial segmentation method that minimizes the sum of intra-segment prediction variance. Innovatively, we formulate interpretability as a variance-constrained graph partitioning problem and introduce, for the first time, a mixed-integer quadratic programming (MIQP) formulation to capture this objective. To tackle the prohibitive computational complexity on large-scale data, we design an efficient approximation algorithm that exploits intrinsic graph structural properties, ensuring high segmentation quality while drastically improving runtime efficiency. Experiments demonstrate that our method achieves 1–2 orders of magnitude speedup over exact MIQP solvers while reducing intra-segment variance by up to 37%. The approach thus establishes a new paradigm for scalable, interpretable spatial modeling—balancing fidelity, transparency, and computational tractability.

0 citationsRead paper

Learning Survival Models with Right-Censored Reporting Delays

Oct 05, 2025

In insurance analytics, risk assessment for newly enrolled populations is hindered by two interrelated challenges: reporting delays (inducing right-censoring) and administrative constraints limiting follow-up duration—both rendering true event times unobservable. To address this, we propose a parametric proportional hazards model that jointly models event occurrence time and reporting delay time. We introduce latent variables representing the underlying event status and perform marginalization over these latent states to enable valid inference. A two-stage estimation framework based on the EM algorithm is developed, ensuring asymptotic consistency and numerical stability. Our method substantially improves both timeliness and accuracy of risk prediction for new cohorts. Extensive simulations and empirical analysis demonstrate rapid convergence, robust performance, and practical utility. The approach establishes a novel, interpretable, and implementable paradigm for survival analysis under reporting delay.

0 citationsRead paper

Recent publications

Latest Papers

Surrogate Graph Partitioning for Spatial Prediction

Oct 09, 2025

This paper addresses the poor interpretability of black-box models in spatial prediction. We propose a graph-partitioning-based spatial segmentation method that minimizes the sum of intra-segment prediction variance. Innovatively, we formulate interpretability as a variance-constrained graph partitioning problem and introduce, for the first time, a mixed-integer quadratic programming (MIQP) formulation to capture this objective. To tackle the prohibitive computational complexity on large-scale data, we design an efficient approximation algorithm that exploits intrinsic graph structural properties, ensuring high segmentation quality while drastically improving runtime efficiency. Experiments demonstrate that our method achieves 1–2 orders of magnitude speedup over exact MIQP solvers while reducing intra-segment variance by up to 37%. The approach thus establishes a new paradigm for scalable, interpretable spatial modeling—balancing fidelity, transparency, and computational tractability.

0 citationsRead paper

Learning Survival Models with Right-Censored Reporting Delays

Oct 05, 2025

In insurance analytics, risk assessment for newly enrolled populations is hindered by two interrelated challenges: reporting delays (inducing right-censoring) and administrative constraints limiting follow-up duration—both rendering true event times unobservable. To address this, we propose a parametric proportional hazards model that jointly models event occurrence time and reporting delay time. We introduce latent variables representing the underlying event status and perform marginalization over these latent states to enable valid inference. A two-stage estimation framework based on the EM algorithm is developed, ensuring asymptotic consistency and numerical stability. Our method substantially improves both timeliness and accuracy of risk prediction for new cohorts. Extensive simulations and empirical analysis demonstrate rapid convergence, robust performance, and practical utility. The approach establishes a novel, interpretable, and implementable paradigm for survival analysis under reporting delay.

0 citationsRead paper