论文部分内容阅读
针对船舶溢油事故的赔偿问题,基于径向基神经网络(RBF)和GA-SVM两种方法,以国际油污基金公约所承认的著名船舶油污事故损害赔偿案例作为学习样本和检验样本,通过建立基准模型和3个对比方案,比较两种方法在赔偿评估中的优劣。其中基准模型为GA-SVM模型,以前16个赔偿案例作为学习样本,后3个赔偿案例作为检验样本,以模糊定量溢油量、油种比重、海况、环境敏感度、清污情况5个主要影响因素作为SVM的输入特征向量,赔偿额作为输出量。3个对比模型为径向基神经网络模型,其中方案1与基准模型基本设定相同;方案2在方案1的基础上,以溢油量、油种比重、风速、波高、涌高、视距(能见度)、环境敏感度、油膜扩散面积、污染海岸线长度、清污情况10个影响因素作为RBF的输入特征向量;方案3以方案2为基准,使用前12个数据为学习样本,后7个为检验样本进行建模分析。结果表明:GA-SVM模型在预测中具有较高的准确率;而RBF在理论上虽占优势,但预测结果不理想,并对其可能的原因作了相应的分析。
Aiming at the problem of compensation for ship oil spill accidents, based on RBF neural network and GA-SVM, using the famous oil pollution accident compensation case recognized by the International Oil Fund Convention as a learning sample and a test sample, Benchmark model and three comparison programs to compare the advantages and disadvantages of the two methods in compensation assessment. Among them, the benchmark model is GA-SVM model, the former 16 compensation cases as learning samples and the latter 3 compensation cases as testing samples, with fuzzy quantitative oil spill amount, oil proportion, sea condition, environmental sensitivity and cleanup conditions as the five main The influencing factors are used as the input eigenvectors of the SVM, with the compensation as the output. The three comparison models are radial basis neural network model, in which scheme 1 is basically the same as the basic model; scheme 2 is based on scheme 1, taking the oil spill amount, specific gravity of oil species, wind speed, wave height, (Visibility), environmental sensitivity, oil film diffusion area, length of polluted coastline, and pollution removal as the input eigenvectors of RBF. In Scheme 3, the first 12 data were used as learning samples and the last 7 For the test sample modeling analysis. The results show that the GA-SVM model has high accuracy in prediction. However, although RBF is predominant in theory, the prediction result is not satisfactory and the possible causes are analyzed accordingly.