纳斯达克半导体行业股指统计特性及其神经网络预测技术研究

金融时间序列统计特性和神经网络预测研究对于掌握金融市场发展规律,并指导长期或短期投资行为具有重要意义。采用经验模态分解(EMD)、时间内禀相关分析(TDIC)和Hilbert谱分析等方法对纳斯达克半导体行业股指进行了尺度统计分析,并利用先验的神经网络对纳斯达克半导体行业股指进行了预测。统计分析发现,各阶本征模态函数(IMF)呈现一定的周期性,能谱分析的结果显示半导体行业股具有统计行为;利用先验的神经网络对半导体股指进行预测,发现半导体行业股指将会在未来一段时间内保持振荡趋势,不同的反向传播(BP)神经网络预测模型可以有效应对半导体行业长期和短期投资方案,可为投资者提供有效的借鉴。

Study on Statistical Characteristics Analysis and Neural Network Prediction Method of NASDAQ Semiconductor Industry Index

The research on statistical characteristics of financial time series and neural network prediction is of great significance for grasping the law of financial market development and guiding long-term or short-term investment behavior. The NASDAQ semiconductor index was investigated by Empirical Mode Decomposition (EMD), Time-Dependent Intrinsic Correlation (TDIC) and Hilbert spectrum analysis, the NASDAQ semiconductor industry index was predicted by the priori-tested BP neural network method. Statistical analysis showed that IMFs presented certain periodicity. The result of the spectrum analysis showed that statistics behavior could be found in the semiconductor industry index. Eventually, it was found that the semiconductor industry index would keep fluctuating in the near future. Different BP neural network methods could be taken to direct the long-term investigation and short-term investigation. The BP neural network method could provide effective reference to investigators.