Good work! A few questions after reading the paper

[gonghaozhang]

I saw the training data contains the NASDAQ and NYSE 1 day frequency data (which I assume to have lower SNR ratio than other?).

1. Do these data positively or negatively affect the performance of the model?
2. How the model perform in iid or ood stock market? Do you have some experiments on that?

[qiu69]

Hello, thank you for your interest and questions:

1. From the results shown in https://github.com/decisionintelligence/DUET/blob/main/figures/other_results.jpg, our algorithm performs exceptionally well on the NASDAQ and NYSE datasets, which have positively influenced the evaluation.

2. Regarding the second question, we did not conduct additional experiments specifically on the stock market. However, among the 25 multivariate datasets we evaluated, some are related to stock data.

3. Additionally, we have released a time series analysis leaderboard: https://decisionintelligence.github.io/OpenTS/. You can find more performance details of various algorithms there, and we hope it will be helpful to you.

Thank you again for your question!

[qiu69]

Hi, If you have any further questions, please feel free to report them to us. If the problem has been resolved, please close the issue.