Feature Request: Add Support for Handling Missing Values

[jotelu]

TimeMixer shows impressive performance on various forecasting tasks. However, it would be helpful to have built-in support for handling missing values in the input time series. This would make the model more robust for real-world applications where data is often incomplete.

Suggested Implementation

• Add an option to specify a masking mechanism for missing values
• Consider implementing an imputation module before the PDM block
• Include a parameter to control how missing values affect the decomposition process

Priority

Medium

Additional Context

Many real-world time series datasets contain missing observations, and the current implementation seems to assume complete data availability.

[kevinliu2000]

Thank you so much for your interest in our work! We truly appreciate your suggestions and are carefully considering your request. As I've recently taken over the project, and with several issues to address, it might take some time to test your needs. However, your input is invaluable to us, and we sincerely thank you for your patience and understanding.