[Summary] the error transmits nerve network reversely (BPNN) as a result of advantageous nonlinear data processing function and stronger learn ability and be applied widely at telegraphic business forecast in the center. However, nerve network often is existing convergence grows to wait for blemish and affect its to forecast the result at time of local and optimum solution, study, and genetic algorithm (GA) is a kind of overall situation the search that find actor is algorithmic, can overcome afore-mentioned limitation effectively. The article is aimed at the main factor that affects telegraphic business income, rise BP nerve network and genetic and algorithmic organic union, built model of corresponding genetic nerve network to be used at telegraphic business income to forecast, use real data to undertake effect test and verify. The experiment makes clear, should forecast a model to have very strong study ability and from adaptability, its forecast model of network of nerve of result excel BP, and have good extensive to turn a sex. [Keyword] income of business of telecommunication of genetic nerve network forecasts 　 　 one, the premise condition that foreword 　 telegraphic business is forecasted is program of construction of communication network grading, also be program period telecommunication at the same time portfolio and income are estimated one of requirement. The choice that forecasts a method matters to the implementation that forecasts an end and the exact rate that calculate a result directly. In recent years technology of artificial nerve network and genetic algorithm get those who forecast scientific worker take seriously gradually, the error travels reversely (BP) nerve network had received wide application in forecasting a domain, be a kind by many nerve yuan the structure of administrative levels network that regulation joins and forms with some kind, its fundamental is these nerve yuan between " mutual cooperation " , it has a lot of advantages, to not complete letter, have good adaptability; to be opposite nonlinear the study that inputs output relation has advantage more, the capability that its describe a problem is very strong. But BP algorithm is a kind of study method that is based on error function gradient to drop, rate of learning process astringent is slower; Next, BP nerve network trains in the begining initiative power worth is random given, this also can have great effect to the training effect of the network, bring about a network to be immersed in even local most bit. Genetic algorithm (GA) has very good overall situation to search ability, can seek the optimum solution of the problem with random means from the meaning of probability. But on the other hand, easy generation is early in genetic and algorithmic applying phenomenon, the local ability that find actor is poorer, and the union of genetic algorithm and nerve network can produce respective advantage. In the training that the article studies to use genetic nerve network to use telegraphic business, the result demonstrates this method is but of effect. Article other structure arranges as follows: The 2nd part passes the structure of nerve network model and algorithm introduction for involuntary discharge of urine; Number of the 3rd share occupies origin and analysis of result of a case of a physically strong patient running a high fever or suffering from such disorders as stasis of blood; It is article conclusion finally. 2, structure of model of genetic nerve network and algorithmic 　 　 1. 　 of structure of model of genetic nerve network is multilayer and forward nerve network is a kind of when be applied generally strong study system in economic domain, construction of system is simple easily process designing. In its in specific applying, the most important is affirmatory network structure above all, and the key of network structure depends on implicit layer and its knot check the number. Consider to make clear, will tell to learning any function, implicit the layer is enough. Because this one three-layer is forward,nerve network can approach random nonlinear function. Advanced to the middle of nerve network structure, if use an error to transmit algorithm reversely (Back Propagation, BP) will undertake study to the authority coefficient of network structure, that is the model of BP nerve network that we say normally. The structure of model of genetic nerve network in this research is built namely in a three-layer forward over nerve network foundation, genetic algorithm and nerve network organic ground links a kind of when rise mixture model. The particular structure of the network is like cent of this network of 　 of graph 1 to be three-layer: Ground floor is input layer, mutual N node; The 2nd is implicit layer, mutual M node; three-layer is output layer, have a node. The target function of the network is, the Y in type is actual output, y by a definite date hopes to output, ep is error squared function. 2. The algorithmic 　 of model of genetic nerve network is among this model algorithm, a kind of improvement heredity is algorithmic, the power that is put forward to optimize model structure is worth coefficient. The characteristic that this algorithm uses genetic algorithm to be good at discovering optimum solution extent above all at the same time, find out the best and initiative value of network parameter, next the optimum solution space that the capacity finding actor of recycle BP algorithm will come search model parameter. Algorithmic and concrete move is as follows: Measure 1: Initialization 　 optimizes nerve network with genetic algorithm, basically be the nerve in optimizing nerve network yuan the join between counterpoises, initialization is planted group of P(t) . Because the join authority of the network is real, because this algorithm adopts real encode plan, avoid weight pace to take change. The network is implicit layer move function is Sigmoid function. In encode process, with the droit of nerve network value and threshold value serve as the gene of chromosome, each gene comprise chromosome vector V=[v1, ... Vk, ... VL] , VK is the K in chromosome gene. Measure 2: Fitness function value is opposite fitness computation and evaluation 　 basis individual undertake evaluating, to every each body undertakes decipher gets network of nerve of a BP inputs example, computation gives the output error of nerve network to be worth E, choose fitness function(www.my-paper.info my paper)