Investigation of the Performance of Soft Computing Methods in the Hydraulic Evaluation of the Slot Fishway on the Inclined Drop

Abstract

The present research investigates K-Nearest Neighbor (K-NN) and Support Vector Machine (SVM) algorithms for evaluating hydraulic parameters of inclined drops equipped with fishway elements. 395 measurements were analyzed using these two algorithms with a focus on ΔE/E and yd/H. To evaluate algorithm effectiveness, three accuracy measures: R2, RMSE, and the KGE are used for a number of different scenarios. The SVM results show that in the first and second scenarios, respectively, model number 13 yielded RMSE=0.0156, R2=0.973, and KGE=0.961 for the training mode and RMSE=0.0241, R2=0.962, and KGE=0.952 for the testing mode. The K-NN algorithm with 75% of the total data employed for the initial sample size gave the best results with RMSE=0.0121, R2=0.986, and KGE=0.975 for the first scenario and RMSE= 0.0123, R2= 0.986, and KGE= 0.975 for the second scenario. Also, the results stated that the RBF kernel function is superior among the kernel functions of the SVM algorithm and the Cityblock distance measurement method among other methods of the K-NN algorithm. On the other hand, the K-NN method was superior to the SVM method.

Keywords

• fishway elements
• inclined drop
• k-nn algorithm
• svm algorithm