Method of immersion of a problem of comparison financial conditions of the enterprises in an expert cover in a class algorithms of artificial intelligence

The financial condition of the enterprise can be estimated by a set of characteristics (solvency and liquidity, structure of the capital, profitability, etc.). The part of financial coefficients is low-informative, and other part contains the interconnected sizes. Therefore for elimination of ambiguity we will pass to the generalized indicators – rating numbers, and as the main means of research it is offered to use the theory of expert systems. As characteristic of the modern theory of expert systems it is necessary to consider application of intellectual ways of data processing of data mining, or simply data mining. The method of immersion of a problem of comparison of a financial condition of economic objects in an expert cover in a class of systems of artificial intelligence is offered (algorithms of a method of the analysis of hierarchies, contiguity leaning of a neural network, algorithm of training with function of activation softmax). The generalized indicator of structure of the capital in the form of rating number is entered and the sign (factorial) space for seven concrete enterprises is created. Quantitative signs (financial coefficients of structure of the capital) are allocated and their normalization by rules of the theory of expert systems is carried out. To the received set of the generalized indicators the method of the analysis of hierarchies is applied: on the basis of a linguistic scale of T. Saaty the ranks of signs reflecting the relative importance of various financial coefficients are defined and the matrix of pair comparisons is constructed. The vector of priority signs on the basis of the solution of the equation for own numbers and own vectors of the mentioned matrix is calculated. As a result the visualization of the received results which has allowed to eliminate difficulties of interpretation of small and negative values of the generalized indicator is carried out. The neural network with contiguity leaning and function of activation softmax is applied to further smoothing of dispersion of indicators. Application of this method allows to facilitate considerably problems of interpretation of results of comparison of a financial condition of various enterprises.