The Neural Modeling for the Assessment of Hazardous Hydro Meteorological Phenomena Including the Case of the Providing a Sustainable Work (Reliability) on Transport

Abstract

The article presents a comprehensive analysis of using artificial neural networks and neural modeling to assess hazardous hydro-meteorological phenomena. The techniques of the undertaken studies are the elements of the mathematic modeling, neural modeling based on the integrated environment MATLAB R2017a, the regression analysis. According to the survey, the conclusion has been made on the expedience of applying the neural modeling to analyze big bulks of data for the assessment of hazardous hydro-meteorological phenomena (hydrological and agro-meteorological inclusive); the geometry of the neural model was selected; the regression analysis of the predicted results and the target outputs was carried out, the regression lines equations and the correlation coefficients were obtained to assess the forecast the neural model utilization under the obtained architecture of the neural network, various training functions, and transfer functions. According to the performed calculations, it was revealed that using the neural network for neural models to assess hazardous hydro-meteorological phenomena (hydrological and agro-meteorological ones included) in the given range of sizes, the normalization of an input, target results in the lesser value of the MSE. In this case, it is appropriate to use the Levenberg-Marquardt algorithm and the transfer function – hyperbolic tangent.