Under the background of global climate change, the rising sea level and the massive emission of greenhouse gases aggravate the loss of typhoon storm surge. It is of great practical significance to build an accurate loss assessment model for Marine disaster prevention and mitigation.In this paper, 50 groups of typhoon storm surges in Guangdong Province from 1995 to 2020 were selected for the experiment. Based on the theory of climate change and risk assessment, the index system of typhoon storm surges loss assessment was established. Principal component analysis was used to screen input factors, and RBF neural network and support vector machine for regression model were used to evaluate the direct economic losses and the affected area of mariculture. By comparing the results of evaluation models with and without climate change indicators, it is verified that climate change is an important factor affecting disaster losses.Entropy weight method combined with RBF neural network and support vector machine for regression model is used to evaluate direct economic losses. Compared with single model, the combined model has better prediction accuracy, which provides a reasonable and effective loss assessment method for disaster prevention and reduction.