Agent-based Stock Index Dynamic Forecasting Model And Analysis

The financial market is a complex and changeable dynamical system,and the researches show that the price fluctuation in the market presents a large number of unusual stylized facts,such as volatility aggregation,power function,Leptokurtosis,long-range relativity,and so on,and the research of stock market volatility becomes a hotspot in academia.Based on fractal theory,this paper constructs the dynamic forecasting model of stock index price on the Sierpinski triangle fractal carpet,and realizes the simulation process through computer programming,and explores the similarity of the simulated data and real data in corresponding statistic regularity,and forecasts the real market.Fractal has been widely used in financial markets,such as using fractal theory to express the invariant regularity of the scale for the financial data,and according to the multifractal theory,we can divide the financial data into different regions based the complexity and learn its internal structure,and so on.In this paper,we use the research results of the fractal lattice of the square to simulate the fluctuation of stock index price based on the Sierpinski triangle carpet.According to the logarithmic return time series of the simulated index,the Leptokurtosis of the financial series is analyzed with the skewness and kurtosis.we explore the complexity with CMSE and analyze the correlation between stock indexes.The result shows that the data simulated by the Agent-Based Stock Index Dynamic Forecasting Model is similar to the actual data in the mentioned statistic regularity above.The model created by Sierpinski triangle is efficient.Based on the historical data,we construct the model to predict the short-term stock index price and compare with the random walk model.Through the error analysis,it shows that the model is valuable in the market forecast.