【摘要】 金融市场是一个复杂、演化、非线性的动态变化的系统.金融数据往往带有噪声,非平稳且时常是混沌的.本文基于时序数据的先验知识——近期数据对于预测未来走势提供了更多的信息,对于传统的支持向量机的回归模型做出了一定的改进,即对于近期的数据预测错误施以更严重的惩罚,构建了改进的支持向量回归机模型.使用该改进模型对中国股票市场指数时间序列进行了预测,结果显示,本文改进的模型较之传统的支持向量回归机模型和神经网络模型有较好的预测效果.更多还原

【Abstract】 Financial market is a complex and nonlinear dynamic system.Financial data is usually noisy,non-stationary and sometimes in chaos.There exists prior knowledge that recent data provides more information for prediction of trends.This paper modifies traditional support vector regression model based on the prior knowledge by penalizing more heavily on recent prediction error.This paper uses the modified support vector regression model to predict Chinese stock market trends.Experimental results show that our model performs better than traditional support vector regression model and neural network model.更多还原