[D] Feature engineering for timeseries datasets

I have 2 timeseries datasets with different variables/features and timestamps. I have done some feature engineering independently on both.

Now, I want to regress on both datasets, as I believe they are related one dataset to the other. However, when I merge both datasets on the timeindex, some rows belong to dataset1 and some to dataset2. As a consequence, all rows have Nan values.

Question:
1. How can I do to manager to fit a linear regression and predict a …

dataset datasets engineering feature feature engineering features however machinelearning merge timeseries variables