Comparative Performance Analysis of Levenberg-Marquardt, Bayesian Regularization and Scaled Conjugate Gradient for the Prediction of flash Floods
Severe flash floods are the root cause of the increased death toll of humans, cattle, and devastation of infrastructure in various countries. Flash floods can be considered as one of the most horrible unpredictable disasters in the world. Floods may cause due to the severe precipitation velocity, cloud to ground flashes, and melting of debris in the ocean. Diversified methodologies were adopted to detect the flash floods rapidly and timely. Almost dozens of sensors have been used to detect all the evaluation parameters concerned to the flash floods like upstream level, precipitation intensity, flood magnitude, run-off velocity, the color of the water, precipitation velocity, pressure, temperature, wind speed, wave pattern and cloud to ground (CG flashes). Transducers and sensing elements produce false alarms or irrelevant information as well due to the inadequate algorithms. False alarm rate is the cause of bad forecasting of floods due to the incompetent algorithms. In this research paper comparative analysis of Bayesian regularization, Levenberg Marquardt, and scaled conjugate gradient-based neural network learning has been performed to determine the best learning algorithm for solving optimization issues with less false alarm rate. Results showed that Bayesian regularization worked better than the other learning algorithms in terms of better fitness, regression value, mean square error in fewer epochs.
Copyright (c) 2020 Journal of Information Communication Technologies and Robotic Applications
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.