Facing rising global disaster risks, the property insurance industry encounters major challenges. This study builds three models to aid insurers in optimizing decisions. The disaster risk assessment model combines Bayesian networks and catastrophe models, using adaptive algorithms and dynamic weighting to precisely gauge regional underwriting risks. It rates southern China 0.8337 (suitable for insurance) and the US Gulf Coast 0.3653 (unsuitable). The property valuation model integrates decision trees and AHP, with an adaptive feature selection mechanism that adjusts weights based on regional traits. It offers a scientific basis for site selection and resource allocation, predicting 14 of 16 US states are fit for new construction and need infrastructure expansion. The building protection model applies fuzzy comprehensive evaluation and entropy weight methods to assess multi-dimensional indicators, making dynamic adjustments and intelligent learning to optimize protection advice. Of 51 evaluated buildings, it suggests protecting 26 medium-value and 21 high-value ones, with stricter measures for the latter. These models provide scientific decision support for insurers, communities, and developers, effectively tackling the challenges of growing disaster risks.