H. Liu et al. : Effluent Quality Prediction of Papermaking WWTPs Using SEL
FIGURE 7. Prediction results of S NHeff and S NOeff using stacking ensemble learning.
TABLE 5. Comparison of modeling results for COD eff and SS eff .
TABLE 6. Comparison of modeling results for S NHeff and S NOeff .
RMSE value (4.25) and the maximum R 2 (0.72) for COD eff , the minimum RMSE value (0.71) and the maximum R 2 (0.68) for SS eff . For S NHeff , compared with the base-learning algorithms of PLS, SVR, and ANN, the RMSE of SEL is reduced by 63.53%, 49.18%, 48.33%, respectively. For S NOeff , the RMSE of SEL is reduced by 57.14%, 42.31%, 47.37%,
13.97%, 6.18%, and 14.49%, respectively. For SS eff , the RMSE of SEL is reduced by 14.46%, 6.58%, and 10.13%, respectively. In terms of R 2 , the prediction accuracy of SEL is also improved significantly range from 5.88%-26.32%, 6.25%-21.43% for COD eff and SS eff . Meanwhile, SEL also demonstrated its superiority of ensemble learning compared with RF and AdaBoost, specifically with the minimum
180851
VOLUME 8, 2020
Made with FlippingBook flipbook maker