Abstract: To realize the short-term prediction of stock prices, we establish a general and concise prediction model based on Laguerre orthogonal basis neural network model (LOBNN). This paper first corrects the weight solving algorithm of the original model, and then adds a new iterative algorithm to the original algorithm (New LOBNN model). In the second step, we use the ar-garch model to pre-process the original data and obtain some corresponding data results. Finally, the final prediction model is obtained by constructing weights of different models and combining them. In order to verify the effect of the algorithm, in the experimental part, this paper uses the improved model to calculate the real data and compares the prediction accuracy of the improved LOBNN & AR-GARCH model with the previous model. The final result proves that the model not only improves the prediction accuracy, but also shows the effectiveness of the improved model in actual scenes.

Key words: stock price prediction, AR-GARCHmodel, the laguerre orthogonal basis neural network, L-M algorithm

摘要： 为实现对股票价格的短期预测,本文在Laguerre正交基神经网络 (LOBNN)模型的基础上,提出了一种新的组合预测模型来预测短期股价的变化。该模型先通过改进LOBNN模型的权值求解算法,用以增强模型的通用性。接着在其基础上设计新的迭代算法,进一步提高模型的预测精度,进而得到新的LOBNN模型。之后将股价数据分别代入AR-GARCH模型和改进后的LOBNN模型,得到输入数据的两组预测值。最后通过不同的权重来组合两种预测结果,生成最终股价的预测结果。文末的仿真结果表明该组合模型在预测精度与通用性上较原始模型有较大的提升,是一种高效的预测模型。

关键词: 股票预测, AR-GARCH模型, Laguerre正交基神经网络, L-M算法