Abstract: The stock price is mutant, nonlinear and random. Single prediction methods can only describe the stock price segment information defect. This paper proposes a stock price combination forecasting model. Autoregressive moving average is used to forecast the stock price’s linear trend, and then the RBF neural network is used to capture the nonlinear part. The two results are combined to form the stock price forecasting results. The simulation experiment is carried out on Baotou Steel shares（600010）. The results show that, compared with single forecast model, the proposed combination forecasting model can describe stock price change rules more comprehensively and accurately. It improves the forecasting precision of stock price.

Key words: stock price, combination forecasting, neural network, autoregressive integrating moving average

摘要： 针对股票价格的突变性、非线性和随机性，单一预测方法仅能描述股票价格片断信息等缺陷，提出一种股票价格组合预测模型。采用自回归移动平均模型（ARIMA）对股票价格进行预测，捕捉股票价格线性变化趋势。采用RBF神经网络对非线性、随机变化规律进行预测。将两者结果组合得到股票价格预测结果。采用组合模型对包钢股份（600010）股票收盘价进行仿真实验，结果表明，相对于单一预测模型，组合预测模型更加全面、准确刻画了股票价格的变化规律，提高了股票价格预测精度。

关键词: 股票价格, 组合预测, 神经网络, 自回归移动差分模型