基于熵的自适应门限小波去噪股市价格预测

Stock Price Forecasting Based on Adaptive Threshold Wavelet Entropy Denoising

本文针对传统软阈值法小波去噪采用统一门限而引起的过平滑问题，根据熵的特性，在各层自适应调整去噪门限，提出一种改进的小波去噪算法，采用Hurst指数和盒维数作为判决准则抑制过平滑。最后将算法应用于股市价格时间序列去噪，并用BP神经网络对去噪后的深发展A近20年的收盘价格进行了分段预测。仿真表明，本文方法与传统方法相比，误差明显减小，预测结果更为理想。

In this paper, duo to the over-smoothing problems of the traditional so.threshold wavelet de-noising which was caused by uniform threshold.Based on characteristics of entropy in time series of stock price,in this article we proposed a new algorithm to filter out the noise using adaptive threshold.Then we take the Hurst index and the box dimension as the decision threshold to justify the effects.Taking“Shenfazhan A”for example,this article forecasts the closing stock price in the recent 20 years.The simulation result indicates that algorithm reduced errors and has more rational de-noising effects.