The Application Of Forward Neural Network Based On Data Mining For The Time Serial Prediction Of Traffic Flow

Recently, Data Mining is becoming a hot topic. It represents a kind of vogue whatever in science and research field or in application field. Why does Data Mining so popular? First of all, it is that it can bring us virtual value, and it is not so hard to get the value. Secondly, the development of Data Mining is not even in every field. In some fields, there is little research about Data Mining. In some other fields, there is much research, but the research is not relates to application properly. There is also some field, where the application research is more advance than the academic research. Because of all these kinds of facts, there is much more developing space for Data Ming in every field and the innovation of theory and practice is comparatively easy. When a new technology has a large application value and it is also easy to develop, its boom is easy to understand.As the expanding of the Data Mining research in every field, every kind of Data Mining technology develops to deep seat. The distinctive new Data Mining technology is Neural Network. Neural Network is a technology deriving from bionics. There is no mature theory to direct the construction of Neural Network, the researchers always get the best structure by trial. But there does be some good theory about how to train a network. The BP arithmetic is this kind of arithmetic that can be used to train the forward Neural Network. And the improving research on BP arithmetic is very popular now. As the developing of computer technology, people can use Neural Network by several function or several lines of codes and they need not to research the detailed Neural Network theory. This makes people much time to devote their energy to the questions that they want to resolve. So the application of Neural Network is becoming flourished.In traffic field, Neural Network has good advantage, in which the most important is that it can deal with the non-linear problem perfectly in this field. The forward Neural Network is specially good at solving the traffic flow forecasting. As the request of intelligence traffic system is on the way, and as the training speed and predict precision is being higher and higher, the application of Neural Network in traffic field is unshakable the wise choice.Due to the actuality and the trend, and on the base of other people's research result, this paper exploits,innovates and make some work in the aspect as follows:1.On the base of summarizing the development of Data Mining, the paper summarizes the concept,duty,method of Data Mining comprehensively and exhaustively, and predict the developing trend of Data Mining according to abundant excuse.2.The paper summarizes the Neural Network technology, and especially summarizes the realization and improvement of BP arithmetic and brings forward the opinions of mine own about the improvement, on which discusses the construction of a generalized neural network.3.The paper expatiates the relation of Data Mining,Neural Network and traffic flow prediction, so makes the skeleton of this paper more clear.4.According to the imitation experiment, the paper discusses the influence which the structure of Neural Network impacts on the result of traffic flow prediction.5.This paper brings forward a kind of four steps intelligent improvement precept, and applies the precept to the prediction of strategic traffic flow. By imitation, the paper proves the excellent effect of these improvements.6.This paper put forward a new type of generalized Neural Network, and applies it to short-term dynamic traffic flow prediction, which is proved by imitation to have a perfect advantage.