一种新的时间序列分析算法及其在股票预测中的应用

RBF-based time-series forecasting

分析了股票市场高度非线性的特点,给出了一种改进的时间序列分析算法。新算法利用径向基网络来对序列中的历史信息进行非线性组合,从而比基于线性组合的时间序列分析算法的基本模型更能有效地挖掘出序列中历史信息之间的相互作用。新算法还利用改进的遗传算法对径向基函数的中心和宽度进行了全局范围的优化选择,进一步提高了径向基网络的非线性映射能力。运用该算法对股票走势进行了预测,取得了令人满意的效果。

The nonlinear properties of stock information were analyzed, and a novel time-series forecasting algorithm was provided. The new algorithm introduced radial basis functions into the basic ARMA model to explore the interaction among past information, and then chose optimal parameters for RBFs using an improved genetic algorithm. Then, selected stock prices trend was forecast using the new .algorithm and approving results were achieved.