Gpr data regression and clustering by the fuzzy support vector machine and regression

TLDR: In this paper, the problem of determining the depth and radius of a circular pipe along with the soil characteristics is studied, using electromagnetic waves with a fuzzy support vector machine as well as a fuzzySupport vector machine to determine the soil, depth, and dimensions.

Abstract: In this paper, the problem of determining the depth and radius of a circular pipe along with the soil characteristics is studied, using electromagnetic waves with a fuzzy support vector machine as well as a fuzzy support vector machine. To this end, three neural network based fuzzy support vectors are used to determine the soil, depth, and dimensions. Also, using the 2D time domain numerical simulations of electromagnetic field scattering, along with MATLAB software, 1030 data are generated for training as well as neural network verification. Given the fact that for each of the three parameters the nature of the problem is different, separate neural networks are considered with different parameters, thus the number of different data for the network training is considered. In all three cases, the neural network parameters are optimized using genetic algorithm to reduce the error and also reduce the number of support vectors. It should be noted that the objective function of the genetic algorithm consists of two components of the error, as well as the number of membership functions, which can be determined by determining a control parameter. For soil permittivity, the algorithm can accurately predict 93% of permittivities, and it decreases to 89.8 for the pipe depth determination. For diameter it is seen that for 69.3 of the cases the algorithm can correctly classify the pipes.