I'm doing multistep univariate timeseries forecasting with XGBRegressor and MultiOutputRegressor, i have sliced train data with varying inputs ex:[(x,y),(250,60),(329,50)...], is there any way I can train these inputs on MultiOutputRegressor without involving comprimise on quality of data, like padding, filling with stats or random data.
I tried to look into different methods using stats but those alter the data,I need to know if there is a method to train varying slices of input that would be immensly helpful, thanks in advance.
