根据股票市场是非线性动力系统的假设，利用混沌理论对混沌时间序列的分析方法，提出了股票价格预测方法。同时利用重构相空间的嵌入维数和延迟时间分别确定经向基函数模型网络的结构和训练样本对，对实际的股票时间序列预测结果表明，该方法能有效地进行短期预测，并与前馈神经网络模型相比，可得到较好的预测结果，因而在股票时间序列预测中有广泛的实用价值。关 键 词 混沌时间序列; 股票价格; 神经网络; 预测Abstract A method of stock price prediction is presented by hypothesis of stock market being non-linear dynamic system and analyzing method of chaos theory for chaos time series in this paper. Meanwhile, structures of radial basic function (RBF) network and pairs of training samples are determined by embedding dimension and delay time of reconstruct phase space respectively. Predicting results for real world stock time series show that the method is able to do effectively short-term prediction. In comparison with traditional forward feedback neural network (BP), the method can make better predicting performance, thus it can be widely used in stock price prediction.Key words chaotic time series; stock price; neural network; prediction