基于多分辨率分析的股票市场预测模型研究

采用多分辨率分析技术将深证成指收盘数据序列分解为多个子序列,然后采用神经网络技术对每个子序列分别建立预测模型,将各个预测结果叠加后得到最终预测结果.研究首先发现多分辨率技术可以有效提高预测模型的预测精度,表明分析我国股市波动时应该按照不同因素对股市影响大小及周期的差异分别研究,才能更有效分析股市运行状况及对其预测;其次结果表明不同类型神经网络预测模型预测性能差异明显,在选择股市预测模型的神经网络类型时应该注意其学习算法及收敛过程,以便能更好捕获股市变化规律.

The Research of Forecasting Model of Stock Market Based on Multifractal Analysis

The series of Compositional Index of Shenzhen Stock Market was decomposed into several subsequences by multifractal analysis firstly,and the forecasting model for each subsequence was made based on neural network.The forecasting result was obtained by integrating the results of subsequence.The study show the multifractal analysis can improve the performance of forecasting model,which indicts we should pay attention the difference of size and period of different factor in order to forecast the stock market.The performance of different type network model is distinct,which show the learning algorithms and convergence process should be considered in selecting neural network for stock market forecasting.