Matrix Science Mathematic (MSMK)

This study evaluates the performance of various fuzzy membership functions (MFs) in predicting volume and bedload rate using sediment data from a bathymetric survey at Ikpoba Dam. Twelve cases with different membership functions: Gaussian, triangular, trapezoidal, and bell-shape were tested across different epochs. The models were assessed based on Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R-squared (R²) values for both training and testing datasets. The Gaussian membership function (Gaussmf), with 7 membership functions and 200 training epochs, outperformed the others, achieving the lowest RMSE of 0.568 (training) and 0.579 (testing), MAE of 0.437 (training) and 0.445 (testing), and highest R² values of 0.914 (training) and 0.932 (testing) for volume prediction. For bedload rate, it also achieved the lowest RMSE of 0.509 (training) and 0.517 (testing), MAE of 0.391 (training) and 0.397 (testing), and highest R² values of 0.9354 (training) and 0.9496 (testing). In contrast, the Trapezoidal membership function (Trapmf) showed the worst performance with RMSE values of 0.874 (training) and 0.905 (testing), MAE values of 0.652 (training) and 0.677 (testing), and R² values of 0.812 (training) and 0.804 (testing). These results emphasize the significance of membership function selection and training epochs in optimizing fuzzy models for environmental and geospatial applications.