basic principles

 Then again, the thesis introduces the basic principles of the BPNN and GA. It combines both to form several GA optimized BPNN based on http://vozforums.com Real-coded and sets up the primary forecast level of the model according to these. Finally, after discussing the related theories of DS ck mens underwear Evidence Theory and elaborating the feasibility that it applied in the field of moncler down jacket exchange rate forecast, it integrates the subnetwork based on GA-BPNN with DS Evdence Theory and sets up 3L-structure predicting model of the exchange cheap nfl jerseys rate forecast. Making use of the 1994-1997 related history data of China and USA, we make the model by Matlab language and get and analyze the result of the experience. The result of the experience demonstrates that the forecast model can more effectively forecast the exchange rate chi flat iron and have the better ability of the exchange rate forecast than the forecast method of the ANN, although it still has some aspects that we aren’t content with. The design of the model strengthens the http://www.sahmy.com credibility and the accuracy of the result, reduces the indetermination of the solution space of forecast targets, increases the predictability of the target and raises the adaptability of the system. Therefore, it has the theoretical and practical siginificance that DS Evidence Theory is chi hair straightener applied in forecast of exchange rate.The main innovative efforts of the thesis are the follow aspects: 1. It applied DS Evidence Theory, the theoretical achievement in Data Fusion, to the forecast of the exchange rate and brings forward a kind of exchange rate forecast model that is formed by GA-BPNN cooperating with DS Theory.China carries out the system of the floating rate. There are many factors that influence the fluctuation of the exchange rate, such as economic factor, political factor, pdychological factor and http://www.guildwarsguru.com/forum so on. At present, when we predict from the data of the exchange rate using the method of ANN, or ignoring these factors, we will find that there are its inhenrent defects; or using different structural variables, there are many problems in aspects of the train, the astringency and the generalization of the network. After introducing the DS Evidence Theory, we can disassemble the whole parametric space of impact factors to several parametric subspaces and design the respective neural network for each parametric space.