search

blue-yonder / tsfresh

Automatic extraction of relevant features from time series:
http://tsfresh.readthedocs.io
MIT License
8.44k stars 1.22k forks source link

05 Timeseries Forecasting: 1 question and 2 edit suggestions #753

Closed AsiehH closed 4 years ago

AsiehH commented 4 years ago

Hi, First I would write my question:

You mention that for X.loc['2020-07-14'],

"the features in this row were only calculated using the time series values of AAPL up to and including 2015-07-15 and the last 20 days."

How can this be so? To predict the value of the price for today, we use the value of the price of today and tomorrow? Shouldn't this be "up to and including 2015-07-14 and the last 20 days." like the other mentions in the notebook?

Edit suggestion: old

If the feature extraction will return a row with the index (AAPL, 2020-07-14 00:00:00), you now it has been calculated using the AAPL data up and including 2020-07-14 (and 20 days of history).

new

If the feature extraction returns a row with the index (AAPL, 2020-07-14 00:00:00), you know it has been calculated using the AAPL data up and including 2020-07-14 (and 20 days of history).

old

We can now use the extracted features to train a regressor.

new

We can know use the extracted features to train a regressor.

nils-braun commented 4 years ago

Thank you very much for the suggestions! And to your question: yes, that one is clearly a typo!

Should I fix it or do you want to do a PR?

AsiehH commented 4 years ago

You can fix it.