Stock Analysis, Prediction And Decision-making Based On Neural Network And Fuzzy Theory

Stock market is a complex non-linear dynamic system. It is very difficult to develop the inherent rules using the traditional timing prediction technology. Based upon the review of traditional prediction methods, this paper puts forward a system composed of neural network and fuzzy theory aimed at stock market analysis, prediction and decision-making, and an ameliorative method on its function is studied. In training of Back-Propagation neural network, parameter adaptable method which can automatically adjust learning rate and inertia factor is employed in order to avoiding systemic error immersed in a local minimum and accelerating the network's convergence; at the same time the further optimization of the network's structure is introduced, compared with conventional Back-Propagation algorithm, the outcome proves that approach precision and generalization ability of neural network have been great improved. With this measure, we make the neural network simulation grounded on fuzzy parameters aimed at stock price of ShenZhen and ShangHai stock market, and progress useful exploration to fully reflect stock price change rules.In the paper, the way that using 'fussy time sequence' to describe basic factors with influence on stock price is advanced, so we employ Membership function of Fuzzy Set to measure and assay original variable. In a word, the way breaks a new path to further definition of dynamic stock market.Compared with classical stock analysis methods, we take stock prediction parts and decision-making implementation into account and make them close integration, from the utility point, set up an effective analytical-operational system for vast investors.The simulation results of optimizations to structure parameter, algorithm parameter and fuzzy judgement show that generalization ability of network have been great improved. Under limited conditions, right trend rates of stock predictionreach 71.32% and decision-making system demonstrates operational occasion for extended investors with lower risks. Diversified results indicate that using network andfuzzy theory to forecast and make decision stock market is one of the successful experiments.