Optimization of coefficient of blurring of the kernel in nonparametric modelling

A modeling of discrete-continuous processes in space “input-output” variables. Modeling of these processes can be carried out using various parametric and nonparametric. This article deals with modeling using nonparametric methods. This decision was taken in view of the fact that non-parametric theory, in contrast to the parametric theory assumes that the only known qualitative characteristics of the process. The modeling objects often have an unknown and complex structure. Given these facts, the use and development of nonparametric theory continues to be an urgent task of our time. Our results can be used to equipment spacecraft modeling and them managing. When building a model of the object by means of nuclear grade, an important parameter - the coefficient blur kernel. The algorithms optimize the ratio blur kernel, namely the method of enumeration, the flexible polyhedron method and genetic algorithm. As an optimization criterion was selected standard error of the test process models, calculated using the sliding test. It is worth also say that the results will be presented in the optimization parameter vector blur kernel (for each input action), and in the optimization of the overall coefficient on the interaction of all the input. As it turns out, the accuracy of the model to optimize the parameters of a blur kernel is slightly inferior to the accuracy of the model with optimized parameter vector blur kernel, and the calculation of the coefficient of blur kernel runs much faster and, as a consequence, the model will be built. These results can be extremely useful in modeling and managing the rapid flow of information and the changing environment.