摘要
为解决传统隐马尔可夫股价行为预测模型对输入特征序列和隐含状态数目敏感,导致预测结果存在局部最优、误差较大的问题,设计了新的股票因子特征选择方法,包括对因子特征的筛选和特征数据预处理。结合贝叶斯信息规则确定模型最佳隐含状态数目,提出了一种优化股价行为预测性能的PRHMM模型。通过对比支持向量机、ARIMA模型,实验结果证明,所提出的预测算法相对传统预测模型,在股价行为预测中有更好的预测表现。
In order to solve the problem of local optimal and large error caused by the sensitivity of the traditional hidden markov price behavior prediction model to the input feature sequence and the number of hidden states,a novel method of stock factor feature selection is designed,including the selection of the factor and the preprocessing of the feature data. Combining Bayesian Information Criterion to determine the number of best hidden states of model,the PRHMM model is proposed to optimize the prediction performance of stock price behavior. By comparing with support vector machine and ARIMA（ Autoregressive Integrated Moving Average） model,experimental results show that the proposed prediction model shows better prediction performance than traditional prediction models in stock price behavior prediction.
作者
喻永生
谢天异丹
刘畅
郭靖雯
张卫东
Yu Yongsheng1, Xie Tianyidan1, Liu Chang2, Guo Jingwen3 , Zhang Weidong1(1. School of Computer Science and Technology, Southwest University of Science and Technology, Mianyang 621010, China; 2. School of Civil Engineering and Architecture, Southwest University of Science and Technology, Mianyang 621010, China; 3. School of Economics and Management, Southwest University of Science and Technology, Mianyang 621010, Chin)
出处
《信息技术与网络安全》
2018年第8期96-100,共5页
Information Technology and Network Security
关键词
隐马尔可夫模型
特征选择
股票价格预测
金融时间序列
Hidden Markov Model （HMM）
feature selection
stock price forecasting
financial time series
