Chinese stock trend prediction based on multi-feature learning and model fusion
Chinese stock trend prediction based on multi-feature learning and model fusion
复制标题
基于多特征学习和模型融合的中概股走势预测
DOI：
10.1109/smds53860.2021.00013
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
Hui Zhou
中科院分区：
文献类型：
--
作者：
Shanyan Lai;Chunyang Ye;Hongyu Jiang;Hui Zhou
股票走势预测是一个极具吸引力的研究课题，准确预测股市走势可以获得更高的投资回报。不幸的是，股市受到各种事件的影响，导致不断波动。因此，很难准确预测股市的变化趋势。挑战在于如何研究各种相关因素并提取有效信息以做出可靠的预测。但新闻信息的独立性较高。这限制了模型的能力。此外，新闻和时间序列数据的融合会导致维度差异的问题。针对这些挑战，我们提出了一种基于注意力的多特征学习方法TCBiGA来预测中国股票走势。对现实世界数据的大量实验表明，我们的方法可以从一组不同的数据中提取更有效的知识库，从而提高模型的性能。
Stock trend predicting is an extremely attractive research issue since the accurate prediction of the stock market trend can obtain higher investment returns. Unfortunately, the stock market is affected by various events, resulting in constant volatility. As a result, it is difficult to accurately predict the stock market changing trend. The challenges lie in how to study various related factors and extract effective information to make reliable predictions. However, the independence of news information is high. This limits the abilities of the model. In addition, the fusion of news and time series data can cause the problem of dimensional disparity. In response to these challenges, we proposed a multi-feature learning method based on attention, TCBiGA, to predict the Chinese stock trends. Extensive experiments on real-world data show that our method can extract a more effective knowledge base from a set of different data, thereby improving the performance of the model.