Preparing multivariate data with multiple FEAT_DYNAMIC_REAL, etc.

[ncolella-mghpcc]

I have examined #615 #695 #1045 and #1095 but I cannot get this formatted correctly. If I have a dataframe with multiple targets and multiple dynamic variables, such as the one shown below, how can I create a dataset to use for training and prediction for each location?

For instance:

Date	Location	Temperature	Humidity	Wind_Speed	Wind_Direction	Pressure
1/1/2021	San Francisco	58	60	5	N	30.1
1/2/2021	San Francisco	60	70	4	NE	30.1
1/3/2021	San Francisco	47	80	6	N	30.2
1/1/2021	Oakland	55	61	10	NE	30.1
1/2/2021	Oakland	58	71	12	E	30.1
1/3/2021	Oakland	50	81	11	NE	30.1
1/1/2021	Sacramento	51	62	15	NE	30.2
1/2/2021	Sacramento	59	72	16	NE	20.3
1/3/2021	Sacramento	54	82	14	NE	30.2

toy_weather_dataset.csv.txt

Using humidity as a target and the other variables as dynamic/static real/cat features.

(Alternatively, I could have one dataframe for each location - either way is workable.)

@AaronSpieler @lostella @PascalIversen @mbohlkeschneider I am sure you all are tremendously busy, but I would greatly appreciate any thoughts you have if/when you have time! 🙏