scour

https://link.springer.com/article/10.1007/s12145-022-00830-7
Construction of input
Aside from data quality, the proper selection of input parameters has the strongest effect on model performance. We used the Pearson correlation coefficient (r) to construct different combinations of input and output parameters (Table 2), as is common practice in hydrology (Choubin et al. 2018; Khozani et al. 2019; Salih et al. 2020; Yaseen et al. 2016). This approach also allows us to assess the effects of different input combination on the result.


https://www.tandfonline.com/doi/full/10.1080/02626667.2020.1754419
The Pearson correlation coefficient (PAC) was calculated for the following conditioning factors: Q, S, R, Q-1, S-1, R-1 and SSL-1 (where −1 indicates 1 month lag time). The PAC results showed that all input variables had a reasonable influence on the suspended sediment load (Table 5); the most and least important conditioning factors were river discharge (0.76) and rainfall depth (0.17), respectively. This result is in accordance with the results of Kisi et al. (Citation2012), who showed that when Q was removed from the model inputs, the predictive power reduced sharply. However, the result contradicts other work which has showed that both discharge and rainfall depth are the two most important factors for suspended sediment modelling (Jie and Yu Citation2011, Kisi et al. Citation2012). This contradiction is likely because in the present paper the watershed is very large, so the connectivity between rainfall, runoff and suspended sediment loads is low.


https://www.sciencedirect.com/science/article/pii/S0022169421001475
2.3. Input combination
Because some input parameters are more important or relevant than others, it is important to exclude those that hamper the modeling performance without improving the effectiveness of the modeling. To select the best input array, four different input combinations were constructed based on the Pearson correlation coefficient and finally tested in order to find the most effective one. To start with, the parameter with the highest degree of Pearson correlation coefficient (r) was considered as the first input parameter to the model. The assumption is that the parameter with the highest correlation with the output has a better ability to predict the output with higher accuracy. Then, the next parameter with the next highest r value was added to the first input and the selection “input No. 2” was defined. This process continued until the parameter with the lowest r value was added to the combination of input parameters and the selection “input No. 4” was defined (Table 3). The most effective input combination was identified by comparing the effectiveness of each input combination using the root mean square error (RMSE).