【摘要】 福建省海洋经济发展综合指数评价过程中往往存在指标的多重共线性以及回归的非线性问题，这都可能导致模型评价结果无法客观反映实际经济发展规律。将海洋GDP作为被解释变量来替换传统综合指数，运用随机森林算法来减少建模过程中指标的共线性影响，建立了4种多层感知机网络模型来验证回归任务中的非线性关系。结果表明：福建省海洋经济发展水平同构建的子系统指标之间存在显著的非线性关系；随机森林算法同多层感知机结合后构建的回归模型能够很好地拟合福建省海洋经济发展状况，且福建省海洋经济在科技创新和生态环境方面存在明显的短板。更多还原

【Abstract】 In the evaluation process of the comprehensive index of marine economic development in Fujian Province, there are often multicollinearity of indicators and nonlinear problems of regression, which may lead to the model evaluation results not being able to objectively reflect the actual economic development law. The traditional composite index with ocean GDP was replaced as an explanatory variable, the random forest algorithm was used to reduce the collinear influence of the index in the modeling process, and four multilayer perceptron network models to verify the nonlinear relationship in the regression task was established. The results show that there is a significant nonlinear relationship between the development level of marine economy in Fujian Province and the subsystem indicators constructed. The regression model constructed by combining the random forest algorithm with the multilayer perceptron can fit the development of the marine economy in Fujian Province, and there are obvious shortcomings in scientific and technological innovation and ecological environment of the marine economy of Fujian Province.更多还原