Time Series Forecasting Using LightGBM

Calix28

Hi Everyone,

I am just new in using Dataiku and I am seeking for help on my project. I am doing a timeseries forecast based on following scenarios:

1. Forecast Data A based on its Historical Data

2. Forecast Data B based on its Historical Data

3. Forecast Data C based on Historical Data of A, B and C (possible to add forecast data of A and .

4. Forecast Data D based on Historical Data of A, B, C and D (possible to add forecast data of A, B and C).

Any idea how to do this in dataiku?

StanG

Hi,

I'm not sure to fully understand your question, but I think you could solve your problem, for instance point 3. using the Visual ML forecasting feature to forecast data C and use data A and B as external features of your model (https://doc.dataiku.com/dss/latest/machine-learning/time-series-forecasting/settings.html#settings-external-features). Model would train on historical data of A, B and C and then during scoring, you would also use forecast values of A and B to forecast future values of C.