摘要
将波浪理论应用于价格、成交量特征样本序列选取,提出基于小波包能量值聚类特征提取与遗传神经网络相结合的股价预测模型.该模型采用小波包系数单支重构能量值空间分布表征价格波动本质,对能量点进行聚类以降低特征向量维数,将遗传算法与BP网络优势互补用于股价预测.对沪市股票上海汽车(600104)等进行的实证研究结果表明,该模型具有收敛速度快和预测精确度高的特点.
By using wave principle to extract time series characters of stock price and trading volume, the stock price forecasting model based on clustering the feature vectors that reflect the energy change of wavelet packet reconstruction and the genetic algorithm integrated with the back-propagation algorithm is presented. Spatial distribution of energy feature vectors of wavelet packet reconstruction is used to represent the essential feature of price fluctuation. Energy values of wavelet coefficients were clustered to solve the problem of dimensions explosion when the number of input data was large. By combining the genetic algorithm with the back-propagation neural network the case studies of co., Ltd(600104) etc. in the SSE suggest the model have fast convergent speed and high predictive precision.
出处
《哈尔滨工业大学学报》
EI
CAS
CSCD
北大核心
2004年第9期1272-1275,共4页
Journal of Harbin Institute of Technology
关键词
波浪理论
小波包分析
模糊C均值聚类
减法聚类
遗传算法
BP算法
wave principle
analysis of wavelet packet
fuzzy c-means clustering
subtractive clustering
genetic algorithm
back-propagation algorithm
