Variable selection method based on filter-wrapped hybrid framework and its application in wastewater treatment processes

In order to achieve the goal of intelligent management of wastewater treatment plants, it is necessary to monitor the parameters of wastewater treatment in real time. Currently, soft measurement technology based on neural network is widely used to solve the problem of key parameters that are difficult to obtain directly in wastewater treatment processes. Selecting appropriate input variables can reduce the complexity of the neural network topology while improving the prediction accuracy of the soft measurement model. To this end, this paper proposes a variable selection method based on the filter-wrapped hybrid framework (FWHVS), which uses the maximum information coefficient (MIC) index and a radial basis function (RBF) neural network. The proposed method only considers removing redundant and irrelevant variables to optimize the review stage of the SFFS method, thus improving computational efficiency. Finally, common variable selection methods are used as comparison methods to verify the effectiveness of the proposed method using the UCI wastewater treatment dataset.