Agent-Based股指价格动态预测模型构造及分析

发布时间：2018-01-06 12:09

本文关键词：Agent-Based股指价格动态预测模型构造及分析　出处：《北京交通大学》2017年硕士论文　论文类型：学位论文

【摘要】：金融市场是一个复杂多变的动力系统,近年来的研究表明市场中的价格波动呈现出大量不寻常的统计规律性(StylizedFacts),比如波动聚集性、尖峰厚尾、长程相关性等等,对股票市场波动性的研究成为学术界的一个热点。本文以分形理论为基础,创造性的在Sierpinski三角形分形地毯上构建股指价格动态预测模型,通过计算机编程实现仿真过程,探究模拟数据与真实数据在相应统计规律性上的相似性,并对真实市场进行短期预测。分形在金融市场中已经得到了应用广泛,例如分形对金融数据表现出来的标度不变规律性从理论高度上重新进行了表述;多重分形将复杂体系分成许多奇异程度不同的区域研究,分层次了解其内部结构;分形的自相似性与市场的波动中整体和局部的相似性特点吻合等。本文借鉴Sierpinski地毯格点分形的研究成果,在Sierpinski三角形上构建渗流模型模拟股指价格波动。针对模拟的股指对数收益率时间序列,从整体分布的角度分析金融时间序列存在的尖峰厚尾分布特性,运用复合多尺度熵的方法探究其复杂性,借助赖时本征相关性的方法分析股指之间的相关性;实证结果表明,股指价格动态预测模型所模拟出来的数据在上述统计规律性上与实际数据存在一致性,说明利用Sierpinski三角形分形创建的模型是合理的。本文最终根据历史数据创建模型用于预测短期股指价格波动,并以随机游走模型作为对比,对预测数据进行误差分析,直观的说明模型在市场预测中的价值。
[Abstract]:Financial market is a complex and changeable dynamic system. Recent studies show that price volatility in the market presents a large number of unusual statistical laws StylizedFacts. Such as volatility aggregation, peak thick tail, long-term correlation and so on, the research of stock market volatility has become a hot topic in academia. This paper is based on fractal theory. The dynamic prediction model of stock index price is constructed on the Sierpinski triangle fractal carpet creatively, and the simulation process is realized by computer programming. To explore the similarity between simulation data and real data in the corresponding statistical regularity, and to predict the real market in the short term. Fractal has been widely used in the financial market. For example, fractal represents the scale invariant law of financial data from the theoretical height; Multifractal divides the complex system into many regions with different degrees of singularity to understand its internal structure at different levels. The self-similarity of fractal coincides with the characteristics of global and local similarity in market volatility. This paper draws lessons from the research results of Sierpinski carpet lattice fractal. The seepage model is constructed on the Sierpinski triangle to simulate the price fluctuation of stock index. This paper analyzes the distribution characteristics of financial time series from the point of view of overall distribution, uses the method of complex multi-scale entropy to explore its complexity, and analyzes the correlation between stock indexes by means of the method of time-dependent correlation. The empirical results show that the data simulated by the dynamic forecasting model of stock index price are consistent with the actual data on the above statistical regularity. Finally, according to the historical data, the model is used to predict the price fluctuation of short-term stock index, and the random walk model is used as a comparison. Error analysis of forecasting data, intuitionistic explanation of the value of the model in market forecasting.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F830.9;F224

【参考文献】