Code and numerical data for Bellman iteration and policy convergence for finite-horizon renewal dividend problems

Description

This dataset contains Python code and numerical results for dividend optimization over a finite horizon in a renewal risk model. The experiments study first-claim Bellman iteration with entropy regularization, including value and dividend-distribution convergence, comparisons between renewal and Poisson arrivals, dependence on claim age and remaining time, and the effect of the entropy parameter. Verification examples include an explicit benchmark, mesh refinement and Monte Carlo evaluation of a fixed dividend distribution. The deposit contains 34 CSV files, a JSON summary, seven reference figure PDFs, the numerical generator, a verification script, dependency information, file descriptions and a figure/table map. The data are computational outputs rather than observations from an insurance portfolio. Code preparation used AI assistance; responsibility for reviewing the numerical implementation and results remains with the author.

Steps to reproduce

Download and extract the package. Using Python 3.12 and NumPy 2.3.5, run python reproduce.py from the package directory. Regenerated files are written to reproduced_data/. Run python verify_results.py to compare the regenerated CSV files with the archived data. For a shorter check, run python verify_results.py --quick. See README.txt for parameters and explanations, and FIGURE_TABLE_MAP.txt for the data underlying the manuscript figures and tables.

Institutions

• Nankai University

  Tianjin, Tianjin

Categories

Licence