A CNN-SVR Hybrid Prediction Model for Wastewater Index Measurement

Abstract

In order to improve the efficiency and prediction accuracy of wastewater treatment in the printing and dyeing industry, in view of the difficulty of measuring the content of wastewater indicator BOD (Biochemical Oxygen Demand), this paper proposes a Convolutional Neural Network and Support Vector Regression hybrid model of wastewater index content prediction model. Firstly, the input easy-to-measure data is constructed in sequence in the form of a window as a model input. Secondly, CNN is used to extract feature vectors. The resulting feature vectors are constructed in a sequence and used as input data for SVR. Finally, SVR is used for index prediction, and compare with convolutional neural network model and support vector regression model. The actual wastewater treatment plant data in the UCI database is used for experiments. The mean absolute error (MAE) and root mean squared error (RMSE) are used as the evaluation criteria. The experimental results show that the CNN-SVR hybrid model proposed in this paper with higher prediction accuracy.