当前位置: X-MOL 学术Indian J. Geo Mar. Sci. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Use of different modeling approach for sensitivity analysis in predicting the Catch per Unit Effort (CPUE) of fish
Indian Journal of Geo-Marine Sciences ( IF 0.4 ) Pub Date : 2020-11-25
V K Yadav, S Jahageerdar, J Adinarayana

The contribution (Sensitivity analysis) of four variables, namely chlorophyll-a (Chl-a), sea surface temperature (SST), photosynthetically active radiation (PAR) and diffuse attenuation coefficient (Kd_490 or Kd) in predicting the Catch per Unit Effort (CPUE) of fish was evaluated using simple General Linear Model, Generalized Linear Model (GLM), Generalized Additive Model (GAM) and different explanatory methods of Artificial Neural Networks (ANN) technique. The models were assessed for their accuracy in determining the relative importance of the four variables in predicting the CPUE. GAM was an improvement over the General Linear Model, while ANN was found better than GAM. The six explanatory methods which can give the relative contribution or importance of variables were compared using ANN modeling techniques: (i) Connection weights algorithm, (ii) Garson’s algorithm (iii) Partial derivatives (PaD) (iv) Profile method (v) Perturb method, and (vi) Classical stepwise (forward and backward) method. Our results showed that the PaD method, Profile method, Input perturbation (50 % noise), and Connection weight approaches were only consistent in identifying the two most important variables (Chlorophyll-a and Kd) in the network. The distribution of profile plot & partial derivative helped indirectly in finding the other three variables in decreasing order of importance (PAR > fishing hour > SST). It was observed that the significance (sensitivity) of independent variables under GAM and explanatory methods of ANN were similar.

中文翻译:

使用不同的建模方法进行敏感性分析,以预测鱼的单位捕获量(CPUE)

四个变量的贡献(敏感性分析),即叶绿素-a(Chl- a),海面温度(SST),光合有效辐射(PAR)和扩散衰减系数(Kd_490或Kd)在预测鱼的单位捕捞量(CPUE)时使用简单的通用线性模型,广义线性模型(GLM)进行了评估,广义可加模型(GAM)和人工神经网络(ANN)技术的不同解释方法。在确定四个变量在预测CPUE的相对重要性方面的准确性时,对模型进行了评估。GAM是对通用线性模型的改进,而ANN被认为比GAM更好。使用ANN建模技术比较了可以给出变量的相对贡献或重要性的六种解释方法:(i)连接权重算法,(ii)Garson算法(iii)偏导数(PaD)(iv)轮廓方法(v)Perturb方法,以及(vi)经典逐步(向前和向后)方法。我们的结果表明,PaD方法,配置文件方法,输入扰动(50%噪声)和连接权重方法仅在确定两个最重要的变量(叶绿素-a和Kd)。剖面图和偏导数的分布以重要性递减的顺序间接帮助找到了其他三个变量(PAR>钓鱼时间> SST)。据观察,GAM下自变量的显着性(敏感性)和ANN的解释方法相似。
更新日期:2020-11-25
down
wechat
bug