Bug with regression models for the multivariate targets fixed
Current problem: when placing some models for time series forecasting in PrimaryNode, an error occurs in the model's fit stage (for forecast_length > 1).
List of models that fails: xgbreg, adareg, gbr, svr, sgdr List of models that work correctly: linear, lasso, ridge, knnreg, dtreg, treg, rfr
The bug fixed by with involvement of MultiOutputRegressor that wraps the models that can not produce multivariate prediction directly (see https://stackoverflow.com/questions/39540123/muti-output-regression-in-xgboost).
Additional:
- multivariate target disable if it not necessary (to speed-up the fit of the ts models)
- ts compsing example extended