what's the difference between interpolation and imputation？

[shenzekai]

Hi, I am researching the papers related to interpolation, and your GitHub is very helpful to me.
I would like to ask what is the difference between interpolation and imputation? One is the spatial dimension and the other is the temporal dimension?
Thank you.

[jlidw]

According to Wiki, “imputation” is the process of replacing missing data with substituted values; "interpolation" is a process of constructing new data points based on known data points. I think this is the reason why we call "time series imputation" and "spatial interpolation": each time series has its own historical recordings, we perform imputation to fill the missing parts; however, for a new location in space, we don't have any historical recordings/observations (no sensor is installed here), so we need to perform interpolation to infer the value of this new point.

Hope this helps.

[shenzekai]

OK, I think this explanation is perfect.
BTW, your work SSIN is excellent.
I summarized the SIGMOD 2023 papers related to spatial-temporal data and other conferences in Zhihu. Hope they are helpful to you. 时空数据顶会论文总结

[jlidw]

Thanks for your work!

I just found out there is a typo in the SIGMOD 2023 时序&时空数据论文总结, Paper 12 is not mine, lol~

[shenzekai]

Thanks for your correction, this typo has been corrected now.

[shenzekai]

Hi, I found two papers about ST interpolation with codes. Maybe you can add them to your list.

1. 2024[KBS]Inductive and adaptive graph convolution networks equipped with constraint task for spatial–temporal traffic data kriging [Paper] [Code] IAGCN
2. 2024[EDBT]Spatial-temporal Forecasting for Regions without Observations [Paper][Code] STSM
   The second paper is different. It looks like a new question. It not only interpolates value in the past but also forecasts value in the future.