Thistextmainlystudiedunderthemodelframethatinformationblend,theintegrationBPnervenetworktechnique, varioustechniques,carryingonanautotothestockcycletimedata,fast,sciencegroundprocessing,docredibility,accuratesituationthevaluationcompletesestimatedecisionthusofmethod。
This text for the sake of realization according to the estimate model of the BP time fusion, made a thorough research to the blemish of the BP nerve network, put forward lead to correct the PEBP(the phase evaluated BP) model of the method into the study stage evaluation and the abnormal state nerve dollar check, overcome at physically applied function obstacle of the BP nerve network, and take in to verify PEBP function through a great deal of improvement experiment。
This text predicts to time sequence the medium creation problem carried on analysis, pointing out the problem that general time sequence model operability differs in the stock price estimate, combining the thought of the information fusion, putting forward the new time sequence model and wave/frequency time sequence estimate method, and design to correspond of time the sequence data prepare processing of the method of the slice and sample heavy 构 。
That method overcame operability for the general time sequence model to used for the stock estimate to differ of problem, and promise the anti- interference ability of the fusion model。
In addition this text leads to go into of the risk property combination theories, become the theories foundation that blends the situation in the model valuation realization, make the system have to select a stock and carry on the ability that combines an investment, promise the anti- risk ability of the system。
Imitate the experiment enunciation a combination the situation valuation of the theories according to the risk property, risk lowered an investment in the stock the investment the decision。
。