Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Understanding of Trend

See original GitHub issue

I have been impressive so far with the capability of Prophet During my current project, I have some ambiguous feeling about trend and yhat After running m.predict(future) and printing out the future dataframe, I realised there are several columns.

I am concerned about trend and yhat.

What is the difference between trend and yhat?
In the trend graph below, what does each colour points represent? (trend or yhat) (black and blue)
In the trend graph below, my professor is asking about the initial trend value to adjust the trend graph. Is there any variable like that and if it is, will you explain briefly about it?

Thank you for your help

Issue Analytics

State:
Created 6 years ago
Comments:8 (4 by maintainers)

Top GitHub Comments

5reactions

BramVanRensbergencommented, Jan 19, 2018

Regarding your first question: yhat is the actual forecast, trend is one of the possible components of that forecast. If you have no seasonality nor additional regressors, I imagine they are identical; but if you add some seasonality and an additional regressor X, for each week yhat would be the sum of trend, seasonality, and contribution of the regressor.

In your graph, the black dots are your actual datapoints (that you provided), dark blue is yhat (possibly identical to trend in your case), light blue reflects the confidence bounds of yhat (by default 80%).

2reactions

blethamcommented, Mar 19, 2018

@tetianakh the details are in the paper preprint here: https://peerj.com/preprints/3190v2.pdf