混沌时间序列在股票价格长期预测中的应用

摘要

For the defect of initial sensitive to chaotic time serial and the cumulative error will enlarge quickly while iterating prediction, this paper proposed an improved support vector machine ( SVM) method to avoid the disadvantages of long-term forecast.The SVM struc-ture was determined based on the reconstruction phase space.To reduce the accumulated er-ror, the existing SVM kernel function was improved to meet the condition which reduced the error.The closing price data of the Bao Steel stock exchange was applied to the improved model to get a better simulation result.%针对混沌时间序列对初值敏感，在迭代预测时累积误差将迅速放大的缺陷，提出一种改进的支持向量机的方法，避免这一弊端进行长期预测。在重构相空间的基础上，确定支持向量机的结构。为减小累积误差，对现有的支持向量机的核函数进行改进，使该核函数满足减小误差的条件。最后把沪市宝钢股份收盘价格的数据应用到改进的模型中，得到较好的仿真结果。