Support Vector regression (scikit learn) giving the same value

time_series
python
scikit-learn

#1

https://notebooks.azure.com/ML-beginner/libraries/1SVR

Not sure, what’s going here. I’m getting the same values for the variable I want to predict peak electrical load.
X is date, y is peak load. I have an idea that I should be using lagged values as my X which should be of the form (no_samples, no_features). I’m trying to predict the load for the last quarter of the year.


#2

Hi,

By itself the SVR does not recognize the parsed dates as a sequential date feature. You need to break it down to year, month and day, even weekday if it helps. Only then the SVR will be able to make sense of the data and train properly.

Regards
Ankit


#3

can you tell me why that is the case?