In order to improve the effectiveness of Fuzzy Synthetic Evaluation (FSE) models, a Parameter Correlation Analysis (PCA) was introduced into the FSE and a case study was carried out in the Naoli River in the Sanjiang ...In order to improve the effectiveness of Fuzzy Synthetic Evaluation (FSE) models, a Parameter Correlation Analysis (PCA) was introduced into the FSE and a case study was carried out in the Naoli River in the Sanjiang Plain, Northeast China. The basic principle of the PCA is that the pairs of parameters which are highly correlated and linear with each other would contribute the same information to an assessment and one of them should be eliminated. The method of the PCA is that a correlation relationship among candidate parameters is examined before the FSE. If there is an apparent nonlinear or curvilinear relationship between two parameters, then both will be retained; if the correlation is significant (p<0.01), and the scatter plot suggests a linear relationship, then one of them will be deleted. However, which one will be deleted? For solving this problem, a sensitivity test was conducted and the higher sensitivity parameters remained. The results indicate that the original data should be preprocessed through the PCA for redundancy and variability. The study shows that introducing the PCA into the FSE can simplify the FSE calculation process greatly, while the results have not been changed much.展开更多
基金Under the auspices of National Natural Science Foundation of China (No. 40830535)Knowledge Innovation Pro-gram of Chinese Academy of Sciences (No.KSCX2-YW-N-46-06)
文摘In order to improve the effectiveness of Fuzzy Synthetic Evaluation (FSE) models, a Parameter Correlation Analysis (PCA) was introduced into the FSE and a case study was carried out in the Naoli River in the Sanjiang Plain, Northeast China. The basic principle of the PCA is that the pairs of parameters which are highly correlated and linear with each other would contribute the same information to an assessment and one of them should be eliminated. The method of the PCA is that a correlation relationship among candidate parameters is examined before the FSE. If there is an apparent nonlinear or curvilinear relationship between two parameters, then both will be retained; if the correlation is significant (p<0.01), and the scatter plot suggests a linear relationship, then one of them will be deleted. However, which one will be deleted? For solving this problem, a sensitivity test was conducted and the higher sensitivity parameters remained. The results indicate that the original data should be preprocessed through the PCA for redundancy and variability. The study shows that introducing the PCA into the FSE can simplify the FSE calculation process greatly, while the results have not been changed much.
