基于后效时间长度的股票价格预测

Stock-Price Forecasting Based on Length of Timeliness

研究股票价格准确预测问题,针对股票走势预测受政治经济变化等多种因素的影响,具有较强的时变性和非线性特性。传统方法多为线性系统的预测方法,不能有效提高预测精度。为准确确定股票走势后效的时间长度,提高预测精度,根据地统计学与支持向量机提出了一种新的股票价格预测方法。先对数据进行平稳化处理并以地统计学分析股票价格数据的结构性,确定后效时间长度,根据后效时间长度确定各样本的拓阶次数,并对数据进行主成分分析,消除各描述特征携带的噪音。最后采用非线性支持向量机对得到的主成分构建模型并预测。以深发展A股与上证A股两个数据集进行仿真,预测精度均明显高于参比模型。仿真结果表明,新方法能准确预测股价走势,且稳定性好,为股价预测领域提供了有效的手段。

The stock price trend prediction is a typical time series with a strong time -varying, and the existing methods have the defect of low prediction accuracy. Effectively determining the length of timeliness is the key issue to improve the prediction accuracy. To improve the prediction accuracy, a new method of stock - price forecasting was proposed based on support vector machines and geostatistics. Firstly, the time series were stabihzed and the structure of the stock prices was analyze. Secondly, the order of data sets was identified based on the length of timeliness. Thirdly, noise information of variables was eliminated by principal component analysis. Lastly, nonlinear model was established by support vector machine. The simulation experiments were based on two actual stock price data sets and the new model has the highest prediction accuracy among the all reference models. The simulation results show that new method can accurately predicts the stock prices and has good application prospects in stock price forecasting.