The method of machine learning based on graphical data

This paper is devoted to a new method of machine learning based on graphical data (named FuzGraph). The method is founded on fuzzy representation of graphical data. Set of samples of similar images on the figure $y=f(x)$ are described as combination of fuzzy functions. These functions are results of fuzzification of density of probability of some geometric parameters of the revealed images. The method is convenient for detection of laws on the graphical data having the stochastic nature. Testing and approbation of the method were passed on the problem of forecasting of the prices of currencies and stocks in a figure “flag”, which is actual for the financial markets.