基于S-V-PSAL混合模型的股票预测研究
Research on Stock Prediction Based on S-V-PSAL Mixed Model

由于股票价格数据具有非平稳、非线性、复杂性高等特点，欲对其进行预测就存在一定的困难，提出了一种基于S-V-PSAL混合模型的预测方法。首先使用奇异谱分析(SSA)对股票历史数据进行一次分解，得到趋势项和噪声项。对于较平稳的趋势项，使用支持向量回归(SVR)模型进行预测；对复杂度依旧很高的噪声项序列，利用变分模态分解(VMD)再次分解，并使用长短期记忆网络和注意力机制(ALSTM)对得到的模态函数(IMFs)和残差序列(res)进行预测。最后将各预测结果重构得到最终结果。文中使用亿纬锂能股票的历史数据对提出的模型进行检验，通过三种评价指标，可以表明提出的模型比其他对比模型得到的预测效果更好，有更高的准确性。
It is difficult to forecast stock price data because of its non-stationary, nonlinear and complex char-acteristics. A prediction method based on S-V-PSAL mixed model is proposed. First, the historical stock data is decomposed by singular spectrum analysis (SSA), and the trend term and noise term are obtained. For stable trend items, support vector regression (SVR) model is used to predict them. The noise sequence is decomposed again by variational mode decomposition (VMD), and the modal function (IMFs) and residual sequence (res) are predicted by long short-term memory network and attention mechanism (ALSTM). Finally, the forecast results are reconstructed to get the final result. In this paper, the historical data of EVE Energy stock is used to test the proposed model. Through three evaluation indexes, it can be shown that the proposed model has better prediction effect and higher accuracy than other comparison models.