search

google-research / timesfm

TimesFM (Time Series Foundation Model) is a pretrained time-series foundation model developed by Google Research for time-series forecasting.
https://research.google/blog/a-decoder-only-foundation-model-for-time-series-forecasting/
Apache License 2.0
3.02k stars 228 forks source link

The problem of making predictions with multiple input features #63

Open zhaokui001 opened 1 month ago

zhaokui001 commented 1 month ago

I have been able to predict the result through single input, such as stock data, and forecast the following data through input of the closing price of the stock. Such prediction effect is not good, so I want to use more characteristic data to predict the closing price, such as trading volume, opening price and so on. However, I also encountered the problem of input dimension, the data input shape was (512,) (my 'context_len' set 512); With multiple inputs, the data input shape is (512, 5) (I used 5 features), and the following error occurs:

ValueError: Input shape must have rank matching LHS. LHS expects rank 2 but actual input has rank 3

So how do I solve this problem?

CoCoNuTeK commented 3 weeks ago

its not supported, i believe currently its only univariate time series with no covariates