Request additional information (val_num, train_data of Exchange-Rate, trafic, and Bitcoin)

[137809404]

I have been exploring your project and noticed some missing details in the code regarding the Exchange-Rate, traffic, and Bitcoin datasets. Specifically, I would appreciate clarification on the validation split ratio and the number of samples used, particularly for the Exchange-Rate dataset. Additionally, could you specify which columns are intended for prediction?

If possible, could you also provide the final datasets of Exchange-Rate, traffic, and Bitcoin used for LoRA fine-tuning? Your help would be greatly appreciated!

Thank you for your time.

[ameliawong1996]

Thank you for your interest in our project and for your detailed questions!

For all datasets (including Electricity, Exchange Rate, Traffic, and Bitcoin), we used a validation split ratio of 20% of the total samples, with the remaining 80% for training. The exact number of samples may vary slightly depending on any preprocessing steps you perform, but the dataset configuration in the code should align closely with these ratios.

Target Variables

• Exchange Rate Dataset: The target variable for prediction is the "AUD/USD" column.
• Bitcoin Dataset: The target variable for prediction is the "VALUE" column with "ID"= "price" .
• Traffic Dataset: The target variable for prediction is the "VOLUME" column.

Exchange Rate Dataset Column Descriptions

To further clarify, here are the detailed descriptions for each column in the Exchange Rate dataset:

• Time: Timestamp indicating the specific time of data collection.
• AUD/USD: The exchange rate of the Australian Dollar to the US Dollar. This is the target prediction variable.
• AUGDP: Gross Domestic Product (GDP) of Australia.
• AUUN: Unemployment rate in Australia.
• AUIR: Interest rate in Australia.
• USAGDP: Gross Domestic Product (GDP) of the United States.
• USAUN: Unemployment rate in the United States.
• USAIR: Interest rate in the United States.

Please feel free to reach out if you have further questions or encounter any issues with the setup. In the meantime, you can generate training datasets based on the examples provided in Appendix 6.2 of the paper.

[137809404]

Thank you so much for your detailed response!