- 分享主题:LSTM for Stock Forcast - 论文标题:Price Forecast with High-Frequency Finance Data: An Autoregressive Recurrent Neural Network Model with Technical Indicators - 论文链接:https://dl.acm.org/doi/pdf/10.1145/3340531.3412738 - 分享人:唐共勇 |
---|
1. Summary
【必写】,推荐使用 grammarly 检查语法问题,尽量参考论文 introduction 的写作方式。需要写出
- 这篇文章解决了什么问题?
- 作者使用了什么方法(不用太细节)来解决了这个问题?
- 你觉得你需要继续去研究哪些概念才会加深你对这篇文章的理解?
This article makes price predictions based on high-frequency trading data, which is different from the traditional prediction of rising and falling. The author completes this task as a regression task. This paper points out that the existing prediction algorithms for high-frequency transactions are too simple, especially since they can not release the performance of LSTM well. In this paper, the task is made into the form of many to many, and some technical indicators are proposed to be added because it is found that LSTM performs best when the window size is 7, and can not make good use of the information longer than 7. Technical indicators are introduced into the model to better capture historical information. The experimental results show that adding technical indicators does improve the performance of the model.
2. 你对于论文的思考
需要写出你自己对于论文的思考,例如优缺点,你的takeaways
优点:
- 应用类文章,针对五分钟的美股数据进行预测推断,指出了LSTM不能有效处理长序列信息,因此将长序列信息转换为特征数据,即加入一部分技术面因子,能够大幅改善预测效果
缺点:
多对一转换为多对多处理