- 分享主题:TCN for Stock - 论文标题:Stock2Vec: A Hybrid Deep Learning Framework for Stock Market Prediction with Representation Learning and Temporal Convolutional Network - 论文链接:https://arxiv.org/pdf/2010.01197v1.pdf - 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 paper mainly proposes a framework for predicting stock prices. The framework mainly includes two parts: stock2vec and prediction module. Stock2vec is a representation method of extracting the relationship between stocks proposed by the author. It learns a kind of stock embedding through some class features and continuous features (market data); In this paper, the prediction module mainly considers two models for time series prediction: LSTM and TCN. LSTM is a kind of neural network with a good effect in processing sequence data. It is a network with RNN structure. TCN can also process sequence data well with the help of causal convolution and hole convolution. The author combines the two prediction models with stock2vec and achieves remarkable results.
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:
1.提出利用一些横截面数据作为类别特征加入训练生成Stock的embedding
缺点:
1.对于Stock2Vec的生成没有明确的算法解释
2.实验的设置不够详细,而且深度学习模型只有LSTM与TCN进行训练,太过单一
3.根据作者文中所述,以及部分实验结果可以发现embedding的结果与行业有着很大依赖性,因此直接加入行业信息可能也会有较好效果
3. 其他
【可选】
网络结构