The financial sector now is most subject to a computerization. Computers and networks provide processing, transfer and storage of the financial data, a manipulation with transactions and forecasting of rates of financial tools. And knowledge of this branch, by virtue of them abstraction, it is easier to transform and process the information, applying directly the mathematical device. It speaks that the science about the finance has been completely thought up by the person and does not describe any natural or physical phenomena for which studying it is necessary to put experiences and experiments, and also carefully to process the data of supervision for creation most the theories approached to the validity. The wide circulation in financial area was received with mathematical methods of processing and the forecasting, constructed on the basis of neural networks.

Now neural network technologies used in the financial analysis, have ceased to be fashionable exotic and do not cause bewilderment of experts. In the world enormous experience of neural network application is saved up, overwhelming majority of the western financial and industrial companies apply neural technology in this or that kind. Now it is possible to speak with confidence, that the barrier of mistrust is broken, encouraging results of the decision of various analytical tasks have appeared.
It is considered, that the following classes of tasks arising in financial area, it is possible to solve effectively with the help of neural networks: 
Forecasting of time lines for a basis neural network methods of processing (a rate of exchange, demand and quotations, futures contracts, etc.) Insurance activity of banks.

Forecasting of bankruptcies for a basis neural systems of recognition.
Definition of rates of bonds and actions of the enterprises with the purpose of an investment of means in these enterprises. 
Application of neural networks to tasks of exchange activity. Forecasting of economic efficiency of financing of economic and innovative projects. Prediction of results of loans.

The fluent analysis of the publications devoted to the decision of problems from area of management by the finance shows great rise of interest to use of neural networks in bank activity. The general rates of growth can be compared only to distribution of personal computers at the end of 80th years, and initiators of rough growth of applications of neural networks, as a rule, are the largest and solid financial organizations, for which it not only a question of prestige - use of the most perspective and high technology information technology - but also an opportunity to diversify the traditional methods in the most various areas of financial activity. Certainly, at all times object of heightened interest and a subject of numerous researches of banks were such areas, as forecasting of financial events, automation of such traditionally heuristic areas, as recognition of the hand-written text (for the analysis of documents and signatures), an estimation of the real estate, an expert estimation of efficiency of investments into this or that project and many other things. Now distribution neural network hybrid expert systems has reached such level, that on separate, to the most successful neural network to expert systems, banks start to code the information and extremely reluctantly distribute scientific results received at it. Popularity of neural networks is explainable from that point of view, that they solve those tasks which always caused interest of banks but which successful decision restrained not enough effective utilization of information and computing resources more effectively. Despite of the big efforts for development of traditional methods of a prediction of the price of actions, creations of bank expert systems for processing the statistical information and forecasting of economic time numbers have been achieved not too impressing successes that speaks a plenty of working factors.

Financial traditional expert systems, in essence, are based on , i.e. use rather rectilinear statistical models. Neural networks on the basis non-linear, do not demand deep understanding of connections between the initial data and results and promise the big advantages before traditional methods. Numerous experiments show, that adaptive networks on a short time interval always predict better, than standard linear models. Now many foreign research centers and financial establishments carry out works on research and application of neural network technologies for the decision of tasks of the economic forecast. The novelty in these areas should be counted a new direction in the analysis and the forecast on a basis neural network the models which are taking into account correlation of economic time numbers with geographical and demographic statistics. The greatest interest, for the majority of consumers neural network hybrid systems, neural networks for forecasting and a prediction of economic time numbers (exchange rates, actions, insurances).