A question about the hotgym application

Hi,guys!I have a problem.I have watched the hotgym example and it’s amazing.But I wonder what if there is a new gym just opened a few days ago,and I want to predict the total energy consumption next month of this gym .How can I achieve that?There are some gyms in the same city,and their energy comsumption data are available just like the data in the hotgym example.

1 Like

The model params used to create the first gym’s model might work for the new gym, but the encoder parameters might need to be changed if the min/max values of the power consumption have breached the bounds of the first gym. Each gym would get its own model, and the models would learn the specific patterns of usage at each gym.

Yes,but now the problem is the new gym opened just a few days ago,can we predict the consumption of the new gym very accurately according to the other gyms’ consumption data?Thanks.

You mean use the same model instance, but change out the data stream and use a different gym all of the sudden? That would produce a lot of anomalies initially, until the model gets used to the new gym’s patterns, but it should work after some time.