基于多信息源的股价趋势预测

发布时间：2018-04-02 19:57

本文选题：多信息源　切入点：股价趋势预测　出处：《计算机科学》2017年10期

【摘要】：股票价格及趋势预测是金融智能研究的热门话题。一直以来,各种各样的信息源被不断尝试用于股价预测,例如基本经济特征、技术指标、网络舆情、财务公告、财政新闻、金融研报等。然而,此类研究大多数只使用一种或两种信息源,使用3种及以上信息源的极为少见。信息源越多意味着能够提供更加丰富的信息内容和更多不同的信息层面。但是由于各种信源的本质不同,其对股票市场的影响程度不同,因此将多种信源融合起来进行股价预测并非易事。此外,多信源也增加了维度灾难的风险。基于信息融合的目的,尝试同时利用基本经济特征、技术指标、网络舆情3种信息源来进行股价预测。具体做法:先对不同类型的信息源数据进行针对性的处理,使其形成统一的数据集,然后使用SVM分类器建立预测模型。实验结果表明,在选用线性核函数和考虑非交易日数据时,使用这3种信源组合的预测模型的预测效果要比使用单一信源或者两两组合的预测效果好。此外,在收集数据时发现,在非交易日(例如周末或停牌期)虽没有买卖但网络舆情剧增。因此,在实验数据中添加了非交易日的舆情情感数据,分类精准度有所提高。研究结果表明,基于多信源融合的股价预测虽然困难,但是在适当地选择特征和针对性地进行数据预处理后会有较好的预测效果。
[Abstract]:Stock price and trend prediction is a hot topic in financial intelligence research.For a long time, various information sources have been used in stock price prediction, such as basic economic characteristics, technical indicators, network public opinion, financial announcement, financial news, financial research and so on.However, most of these studies use only one or two sources of information, and the use of three or more sources is extremely rare.The more information sources are available, the richer the information content and the more different levels of information.However, due to the different nature of various information sources and their different impact on the stock market, it is not easy to combine various information sources to predict stock prices.In addition, multiple sources also increase the risk of dimensional disasters.Based on the purpose of information fusion, this paper tries to use three kinds of information sources, such as basic economic characteristics, technical index and network public opinion, to forecast stock price simultaneously.Concrete measures: firstly, the different types of information source data are processed pertinently to form a unified data set, and then the prediction model is established by using SVM classifier.The experimental results show that the prediction effect of the three sources combination is better than that of single source or pairwise combination when the linear kernel function is selected and the non-trading date is considered.In addition, data collection found that in non-trading days (such as weekend or suspension period) although not bought and sold, but the Internet public opinion surge.Therefore, the non-trading day public sentiment data are added to the experimental data, and the classification accuracy is improved.The results show that the stock price prediction based on multi-source fusion is difficult, but it will have a better prediction effect after proper selection of features and targeted data preprocessing.
【作者单位】：广东工业大学计算机学院;暨南大学信息科学技术学院计算机科学系;
【基金】：广东省自然科学基金(2016A030313084,2016A030313700,2014A030313374) 中央高校基本科研业务费专项资金资助项目(21615438) 广东省科技计划项目(2015B010128007)资助
【分类号】：F832.51;TP18

【相似文献】