基于支持向量机的黄东海富营养化快速评价技术

以发展黄东海富营养化现场快速监测技术为目的,选取有色溶解有机物(CDOM)特征吸收系数a_CDOM(255)、a_CDOM(355)、a_CDOM(455)及能现场实时监测的浊度(Tur)、叶绿素a(Chl a)、溶氧(DO)等水质参数,以TRIX值为参照,利用支持向量机建立了近海富营养化快速评价技术.建立的支持向量机模型最优惩罚参数C=45.3,最优核函数参数g=0.7,对训练集分类准确率为92.5%,交叉验证准确率为91.8%,验证集分类准确率为85.0%.结果表明:基于CDOM吸收系数及DO、Chl a、Tur建立的近海富营养化快速评价技术能够准确的对近海富营养化状态进行评估,可为近海富营养化的现场快速监测提供技术支持.

A support vector machine-basedtechnology for rapidly assessing trophic status of the Yellow Sea and the East China Sea

In this study, wedeveloped a support vector machine-based model for rapidly assessing trophic status of coastal watersby easy-to-measure parameters (a_CDOM(255), a_CDOM(355), a_CDOM(455), turbidity (Tur), chlorophyll a (Chl a) and dissolved oxygen (DO)) with the trophic index (TRIX) serving as a reference.After the optimal penalty parameter C(45.3) and kernel parameter g (0.7) were obtained by Grid Search, the SVM model was established and its classificationaccuracy rate was 92.5% for the training data, 85.0% for the validation dataand 91.8% for the cross-validation. The results indicated that the developed technique could be useful for rapidly assessingthe eutrophication status ofcoastal waters.