Other datasets

[naive-art]

Hi, thank you for your open-source code and work. I will consider citing it in my future research. I noticed that the code only uses public datasets such as ETTh1. Is it feasible to apply the code to other datasets (such as stock and other financial datasets)? What kind of operations would need to be implemented? Thank you.

[lss-1138]

Thank you for your interest in our work!
Yes, it is feasible to apply our code to other datasets, including stock or other financial time series data. Here's how you can do that:

1. Prepare your dataset: Place your dataset (in CSV format) under the ./dataset directory. The first column should be the timestamp, and the remaining columns should represent the variables (i.e., different features in your multivariate time series).
2. Write your run script: You can refer to scripts/SparseTSF/linear/electricity.sh as an example. Set data_name=custom to use the Dataset_Custom class defined in data_provider/data_loader.py. This class automatically splits the dataset into training, validation, and test sets with a default ratio of 7:1:2. You should also adjust other parameters such as data_path_name, enc_in, etc., according to your dataset.
3. Run your script: Once your script is set up, you can execute it to train and evaluate the model on your dataset.

Let us know if you run into any issues or need further assistance!