the huge structure

It can resolve the problem such as the huge structure, the long training cost, weak reasoning ability, bad generalization, low flexibility and so on. And it solved the ANNs data fusion problem divided by using DS Theory forecast level. Taking the http://www.chinalinuxpub.com/bbs advantage of the ck women’s underwears output of ANN, we can compute the basic Possibility assign (BPA). Thus we can resolve the haunting problem that BPA function is not constructed easily in DS Theory.2. It carries out the moncler jackets algorithm of GA optimized BPNN based on real-coded in the primary level of the forecast model and designs two subnetwork of the exchange rate, the direction and the value of the forecast according to the characteristics of the model.In the process of designing the primary level of the model, after analyzing the defects of the traditional BPNN, we put forward several improving measures, and optimize BPNN conjunction parameters http://www.alqimmah.net taking advantage of GA. The GA optimized BPNN based on real- coded overcome some inherent disadvantages of BP Algorithm and increase reliability of each subnetwork in some cheap nfl jerseys degree. For the forecast of the exchange rate, it usually divides into forecasting the direction and the value of the exchange rate. Generally speaking, one model only accordingly finishes the forecast of one quality. In this thesis, through accordingly dealing with the primary level and the frame of discernment, the model can finish the forecast of two qualities in the similar way.3. In designing concept and choosing tools of the model, it successfully integrates some kinds of main forecast methods in the field of the chi hair straightener exchange rate forecast and applies the excellent theoretical achievments of the Artificial Intelligent theory and Data Fusion theory. It makes the model have great practical and theoretical extension.The design of the model follows the basic thoughts of Structured Variable Forecasting method, adopts heterogeneous variable and integrates ANN with DS Evidence Theory in the aspect of the application of theory. In the http://www.purerobbie.com process of designing the primary forecast, we adopt the method of Prediction of Time Series method. Actually Prediction of Time Series method based on exchange rate’s historical data can be joined into the primary level as evidence. In the aspect of tools, we integrate GA with ANN and overcome many defects of using simplex ANN in some degree. In addition, the result of the primary level can apply the political and military news and experts’ experiential knowledge as evidence to the whole forecasting level. In this way, it also applies the forecasting method of Expert Experience to this model. The thesis not only applies the theories above to practice, also extends the range of theory achievements. It has great theoritical and practical significance and makes the model have the great extension.