【摘要】 针对股票市场高燥声、强非线性和不确定性等特点和以往传统神经网络预测方法存在的不足,提出了一种基于支持向量机的股市预测方法。该方法主要运用了支持向量机回归的方法结合滚动时间窗来学习建摸。首先通过把低维输入空间的输入向量映射到高维特征空间,将非线性问题转化为线性,然后在结构风险最小化原则下进行二次规划,并求得最优解,从而建立模型。从仿真实验中可以看到,该方法建立的模型较为准确地预测了600009、000815两只股票的日均价,表现出了较强的泛化能力。更多还原

【Abstract】 The stock market is a noisy,non-linear and uncertain dynamical system,and the traditional forecasting method-Neural Networks has some shortages.For those,a forecasting method of stock market based on support vector machine is proposed .It applies support vector regression with rolling time frame to build model.Firstly,through vectors in input space mapping to high dimensional feature space, the non-linear problem is transformed into linear one.Then the optimization solution is obtained by quadratic programming.Finally the model is built.The experiment shows that the method has a high generalization performance.更多还原