一种基于流特征模式的股市跟踪预测算法

发布时间：2018-01-14 01:24

本文关键词：一种基于流特征模式的股市跟踪预测算法　出处：《计算机科学》2013年12期 　论文类型：期刊论文

【摘要】：由于股市波动的突发性、多变性,且时序数据呈非正态分布,传统的时序预测模型难以有效预测股市。提出了一种基于流特征模式的股市跟踪预测算法(SFM-PG),该算法根据股票之间的相关性构建贝叶斯网络,选取目标股票的马尔科夫毯作为其同辈群体,然后基于同辈群体之间的接近度,给出一种窗口跟踪式预测模型,其通过对同辈群体权重的动态更新进行跟踪式预测,以减少股票数据分布非正态性对预测的影响;进而,使用滑动窗口提取时序数据中的特征并形成流特征,通过与模式知识库的匹配提取流特征模式,并利用与流特征模式对应的知识调整预测结果,以减少由于突变所引入的预测误差。最后,在上证股票板块网络上的实验结果显示了算法的实用性和有效性。
[Abstract]:Due to the sudden volatility of the stock market, variability, and time series data are non-normal distribution. The traditional time series prediction model is difficult to predict the stock market effectively. A stock market tracking prediction algorithm based on the flow feature model is proposed, which constructs Bayesian network according to the correlation between stocks. The Markov blanket of the target stock is selected as the peer group, and then based on the proximity between the peer groups, a window tracking prediction model is presented. In order to reduce the influence of the non-normal distribution of stock data on the prediction, the dynamic update of peer group weight can be used to track the prediction. Then, using sliding window to extract features from time series data and form flow features, the flow feature patterns are extracted by matching the pattern knowledge base, and the prediction results are adjusted by using the knowledge corresponding to the flow feature patterns. Finally, the experimental results on the stock plate network of Shanghai Stock Exchange show the practicability and effectiveness of the algorithm.
【作者单位】：合肥工业大学计算机与信息学院;
【基金】：国家自然科学基金(61175051,61070131,61175033)资助
【分类号】：TP18;F830.91
【正文快照】： 1引言股票价格预测对战略投资组合以及经济发展具有重要的作用,但是,非量化因素会严重影响股票价格的走势,如政治事件、经济情况、投资者的情绪、欺诈等,诸多非量化因素导致股票价格波动的突发性和多变性,股票价格时序数据呈非正态分布[1],使股票价格的精确预测是极具挑战的

【共引文献】