I have a dataset similar to the example below:
It contains some categorical data (e.g. business type, business size, location) and the time-series of energy consumption from Time 1 to Time N with a minute granularity (Cons_T1 ... Cons_TN). I need to forecast the energy consumption of new businesses for which I only have the categorical data (business type etc.) but no historical information on time series of energy consumption.
One method I thought about was: - 1. use the categorical data of the existing businesses to cluster businesses into classes - 2. use the categorical data of the new businesses to assign them to the created classes using classification techniques (e.g. ANN) - 3. forecast energy consumption of the existing businesses - 4. use the average of the forecasts of each class as the estimate of the forecast for the new businesses belonging to that class
However, this method would give me the same estimate for each new business belonging to the same class.
Is there any other method to be used for this type of problem? I would appreciate simple explanations with reference to some python library if possible.