Time series documentation #1896
Conversation
Codecov Report
@@ Coverage Diff @@
## main #1896 +/- ##
=======================================
Coverage 98.78% 98.78%
=======================================
Files 149 149
Lines 16424 16424
=======================================
Hits 16224 16224
Misses 200 200 Continue to review full report at Codecov.
|
tamargrey
force-pushed
the
time-series-documentation
branch
from
13c9490
to
1671a62
Compare
tamargrey
requested review from
rwedge,
thehomebrewnerd,
freddyaboulton and
gsheni
docs/source/guides/time_series.ipynb
Outdated
| "In multi-table datasets, a feature engineering window for a single row in the target DataFrame extends forward in time over observations in child DataFrames starting at the time index and ending when either th cutoff time or last time index is reached. \n", | ||
| "\n", | ||
| "<p style=\"margin:30px\">\n", | ||
| " <img style=\"display:inline; margin-right:50px\" width=100% src=\"../_static/images/multi_table_FE_timeline.png\" alt=\"Featuretools\" />\n", |
There was a problem hiding this comment.
This image is not displaying properly in the readthedocs build.
docs/source/guides/time_series.ipynb
Outdated
| "We also need to determine how far back in time before `t - 7` we can go. Too far back, and we may lose the potency of our recent observations, but too recent, and we may not capture the full spectrum of behaviors displayed by the data. In this example, let's say that we only want to look at 5 days worth of data at a time. We'll call this our `window_length`. \n", | ||
| "\n", | ||
| "<p style=\"margin:30px\">\n", | ||
| " <img style=\"display:inline; margin-right:50px\" width=100% src=\"../_static/images/time_series_FE_timeline.png\" alt=\"Featuretools\" />\n", |
There was a problem hiding this comment.
Image not displaying correctly.
docs/source/guides/time_series.ipynb
Outdated
| "Let's take a look at an actual feature engineering window as we defined with `gap` and `window_length` above. Below is an example of how we can extract many features using the same feature engineering window without exposing our target value.\n", | ||
| "\n", | ||
| "<p style=\"margin:30px\">\n", | ||
| " <img style=\"display:inline; margin-right:50px\" width=100% src=\"../_static/images/window_calculations.png\" alt=\"Featuretools\" />\n", |
There was a problem hiding this comment.
Image not displaying correctly.
docs/source/guides/time_series.ipynb
Outdated
| "id": "a8104f18", | ||
| "metadata": {}, | ||
| "source": [ | ||
| "# Time Series Problems" |
There was a problem hiding this comment.
How do you feel about calling this Feature Engineering for Time Series Problems since that is really the focus of this guide?
There was a problem hiding this comment.
that makes a lot of sense to me!
tamargrey
force-pushed
the
time-series-documentation
branch
from
bd7e7ee
to
fbb19e4
Compare
|
Where did all the github checks go? |
@rwedge not sure! Let me try getting the latest from main and pushing and seeing if that kicks off the CI run |
tamargrey
force-pushed
the
time-series-documentation
branch
from
f5e2ced
to
c9b5755
Compare
docs/source/guides/time_series.ipynb
Outdated
| "source": [ | ||
| "### Rolling Transform Primitives\n", | ||
| "\n", | ||
| "Since we have access to the entire feature engineering window, we can aggregate over that window. Featuretools has several rolling primitives with which we can achieve this. Here, we'll use the `RollingMean` primitives `RollingMin`, setting the `gap` and `window_length` accordingly. Here, the gap is incredibly important, because when the gap is zero, it means the current observation's taret value is present in the window, which exposes our target.\n", |
There was a problem hiding this comment.
Here, we'll use the
RollingMeanprimitivesRollingMin,
awkward phrasing
There was a problem hiding this comment.
changing to "Here, we'll use the RollingMean and RollingMin primitives"
tamargrey
force-pushed
the
time-series-documentation
branch
2 times, most recently
from
f058121
to
4fd5704
Compare
tamargrey
force-pushed
the
time-series-documentation
branch
from
4fd5704
to
775190a
Compare
docs/source/release_notes.rst
Outdated
| @@ -27,7 +29,7 @@ v1.6.0 Feb 17, 2022 | |||
| * Fix URL deserialization file (:pr:`1909`) | |||
|
|
|||
| Thanks to the following people for contributing to this release: | |||
| :user:`jeff-hernandez`, :user:`rwedge`, :user:`thehomebrewnerd` | |||
| :user:`jeff-hernandez`, :user:`rwedge`, :user:`thehomebrewnerd` | |||
There was a problem hiding this comment.
extra whitespace between rwedge and thehomebrewnerd
Adds a separate time series guide for how to do feature engineering for time series problems.
closes #1758