The Intelligent Forecasting Of Stock Based On Neural Network And Genetic Algorithm

This dissertation focuses on middle-term intelligent forecasting of stock. The system involves three parts: filter of input variables; Neural Network forecasting modeling; Optimization of network structure and application.The new ideas are applied during the select of input variables and the optimization of network. The results are effective. Thus in a word: the ideas that are put forward are valuable not only in theory but in application.1. The fuzzy curve method is applied for selecting the simple and effective inputs. Be compared with the other statistic methods, it don't need huge computation and it select not only the most important inputs but remove the most relative variables.2. ANN modeling. The model is single hidden layer, multi-inputs and single output system. The output is the price trend of future six weeks. GA is applied to solve the question of hidden layer crunodes and select the connect coefficients.3. The application of the methods. The paper forecast price of future thirties days of "Dong-feng Stock" and "shanghai Stock Index". The method not only optimize the structure and modulus of ANN, but improve the forecast precision.