I would like to calculate a value based on the number of lines on outage at a given time.

An outage consists of the two endpoints and the voltage of the line:

outage = ["node1", "node2", "120"]

The historical values I have are based on per hour time stamps, so the format I believe I will need to feed the model is like so:

[timestamp, [outage1, outage2, etc...], value]

Do most machine learning libraries accept data in a variable width format like this? Alternatively, the only other thing I can think of is to have a super-wide data set and do a one hot encoding where each column is a node and its value is the voltage (scaled perhaps).


    [timestamp, node1, node2, node3, etc..., value]  
    ["1/1/2019 14:00", 120, 120, 0, etc..., 5]

There are thousands of nodes on this grid. What is the best approach for formatting this data in a way that most models will accept?


1 Answer 1


Looking back at this, I've ended up finding an answer to my own question.

What I am trying to do here is a pivot, which converts categorical data into multiple columns, and then sets the value to that of another column.

Additionally, I've found that it is very common for models to have thousands of columns, if not more.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.