Timeseries forecasting model on a multiindex table

TensorFlow General Discussion
1

Hi!
I want to build a timeseries forecasting model on this sataset I have. The dataset has a column ‘time_period’ that states the timestamps, 2 categorical columns (clientID and country), and 2 Y values. However, there are multiple records in each time period.

for instance,
Time - ID - Country - Y1 - Y2
Jan 1 - C01 - India - 0.1 - 0.2
Jan 1 - C02 - UAE- 0.05 - 0.1
Jan 1 - C02 - India - 0.13 - 0.99
Jan 1 - C02 - UAE - 0.12 - 0.88
Jan 2 - C01 - India - 1.1 - 0.23
Jan 2 - C02 - UAE- 2.05 - 0.13
Jan 2 - C02 - India - 3.13 - 0.29
Jan 2 - C02 - UAE - 0.2312 - 0.288

and so on.

How do I create a timeseries forecasting model on this (LSTM, or any neural network based model.)

3

Hi @Hadi_Nisamudeen,

If you want to forecast both Y1 and Y2 in your model, use a multi-output model. I recommend using LSTM for this scenario but you may need to reshape your data as 3D(samples, timesteps and features).

Kindly refer to this time series tutorial for more information.

Thank You.