A Study On Stock Price Forecast Which Based On Gray System Theory

This study was designed to take advantage of gray system theory of the generalrules of the stock market development, optimized gray model that the GM model thequantitative prediction of the stock to explore.With the stock market continues to develop the law on the stock marketunderstanding gradually deepened to produce a variety of stock forecasting methods andanalytical tools. The stock market is constantly changing, traditional methods becauseof the defects can not be effectively and accurately predict stock price changes, soshould be explored and better effectively predict stock price volatility, complexity andregularity of the method.The core of gray system theory and is based on the gray model, referred to the GMmodel, this study must first establish the mathematical model of the system, and thusthe correlation between the overall function of the system, as well as the variouselements of the system, causality, dynamicthe relationship between the specificquantitative studies. Because of the complexity of stock movements, the theory can notfully meet the needs of the stock price prediction. In order to further improve theprecision of the model GM(1,1) model residual correction. GM(1,1) model theremaining residuals model and the model for stock price prediction.This study adopts the modified model of residual SanAiFu as a research object,first calculated by GM(1,1) model simulation sequence, and calculate the residuals. Themodel requires the original series are non-negative sequence, the researchers give theresiduals plus the absolute value of its minimum value to make it meet the conditions ofnon-negative sequence. Then use the residual error correction model derived predictivevalue after the correction. Poor final residual test, association test and a posteriori test toverify the eligibility of the model.Studies have shown: GM(1,1) model to predict the average relative error is0.01272; use with the residual modified GM(1,1) the average relative error of the predicted correction0.01146. The results show that, with residual modified GM(1,1)model prediction than without residual modified GM(1,1) model simulated better. Stocktrading operation, however, the complexity of diversity and how to more effectivelyoptimize the GM model, provide a more accurate prediction for stock investors buyingand selling operations, will continue to be an important topic for the future.