Research And Application On Regression Analysis Of Water Quality Data

Regression analysis, as a common and important method of data mining, palyed an important role when mining useful information contained in the data. With a large number of experts and scholars to study it, regression analysis theory continuous development and improvement, and in all areas of society have been widely applied and has achieved good results. In the field of aquatic ecosystems also has a growing number of applications. In recent years, a large number of nutrients into water, leading to increasingly high degree of eutrophication and a serious threat to the health of water ecosystem. Chlorophyll-a is the main component of chlorophyll in algae, and its content is closely related to the type and quantity of algae in water body. It is one of the important indicators to characterize the phenomenon of eutrophication and its degree. Therefore, to obtain accurate water chlorophyll-a concentration is important for eutrophication control.Support vector machine(SVM) shows good performance in solving problem of nonlinear and small samples, so it is widely used in various fields. In recent years, many scholars mainly from two aspects to improve the prediction accuracy of the support vector machine, one is to optimize the parameters of support vector machines, two is to improve the approximate kernel function. But these efforts only focus on the improvement of the support vector machine itself and ignore the further extraction of useful information from the data set. Therefore, this article proposed a hybrid regression prediction model based on least squares support vector machine and RBF neural network. This model used the useful information contiand in the error terms that produced in training by support vector machine to modifying the forecasting result.Because the water ecosystem is a dynamic, open and sophisticated systems, water quality data contains a lot of uncertainties, and fuzzy rough set as a powerful mathematical tool plays an important role, so fuzzy rough set theory is widely used in the treatment of water quality data. But in the regression analysis model, the fuzzy rough set theory based on fuzzy similarity relation is sensitive to the outliers in the data when calculating the upper and lower approximation, to solve this problem, this article introduced the concept of soft distance and proposed the soft fuzzy rough set which is suitable for regression analysis, then constructed a robust fuzzy rough regression analysis algorithm.Finally, the regression model applied to the “Three Gorges Reservoir Online Monitoring System”, and the model is stable in the system, providing data support for environment governance of Three Gorges Reservoir, it reflects the practicality of the model.