Predict datetime variable with orange canvas

Question

Since Orange is able to read datetime variables, i tried to build a simple workflow to predict datetime variables via regression. Loading the data etc. is no problem, but when it comes to predict the data, the output of Predictions Widget (in this case with AdaBoost Regression) is something like:

1450396800.000; 1438905600.000; 1438732800.000;

whereas the original datetime is:

2015-12-22; 2015-08-07; 2015-08-05;

What does that mean? Did i make any mistakes? Is there any other way to predict datetime variables in orange?

The workflow does its job for any "common" continuous variable.

Answer 1

It appears those numbers are Unix timestamps. The numbers you quote correspond to the following human-readable dates in ISO format:

Epoch date  Human readable date (GMT) 
1450396800  2015-12-18 00:00:00
1438905600  2015-08-07 00:00:00
1438732800  2015-08-05 00:00:00

You can use an online epoch converter to confirm.

Apparently, Predictions widget treats datetime variables as continuous variables (which is ok) but fails to present them nicely. You should probably submit a bug report about it.