The Forecast Model Of Futures Based On RBF Neural Network

To forecast the price of futures is a hot field of futures studies. Investors hope that it can help them avoid risk and make more profit from futures. In this thesis, the writer uses neural network technology to forecast the futures' price, building a forecasting model called The Forecast Model of Futures Based on RBF neural network with the background of Dalian Commodity Exchange, hoping to serve investors better.The research works mainly on three aspects: firstly, it combines the regression tree algorithm and RBF neural network, which uses traditional regression tree algorithm to initialize the centers of RBF neural network's hidden neural cells. And then it compares it's capability with other traditional RBF neural network's learning algorithms. Secondly, it makes the forecasting model true and explains how to select samples, the rules for forecasting and errors control etc. It is really an integrated platform that we can use it to collect raw data, generate samples and forecast. Thirdly, it presents us experimental data. Comparing to the factual data collected, it provides us the analysis of errors and makes sure that the model is usable and reasonable from the angle of promoting business.Finally it tells us that the RBF neural network's learning algorithm based on regression tree is super than other traditionals under a certain circumstance which the number of samples is not very big and the sample's dimensionality is not very big too. At the meantime the forecasting model is reasonable and in effect on price predicting in short time that can give some useful decision making information to investors.