Smoothing in charts, useless?
The Line > Smoothing option gives some weird artifacts with loops and multiple values of Y for a single value of X, which I found hilarious :
The same line chart without smoothing.
I'm guessing it's some kind of bezier curve but doing something with the control points that I don't understand.
Does anybody know for sure that kind of smoothing are they applying here?
and isn't there a more appropriate form of smoothing/ curve-fitting for this context, like regular spline smoothing?
Any way I created a Product Idea https://community.dataiku.com/t5/Product-Ideas/Implement-more-sensible-line-smoothing-algorithms/idi-p/37056 to ask for support of other "smoothing" techniques
In case anyone want to reproduce it, I attached the dataiku_smoothing.csv file to this post. You just need to create a project > Dataset > New dataset > Upload your files > dataiku_smoothing.csv
Then navigate to the Dataset > Charts and add
* Show Y : quantity (AVG)
* By X: order_date (automatic)
The Lines > Smoothing (is on by default)
The result is
Answers
-
Alexandru Dataiker, Dataiku DSS Core Designer, Dataiku DSS ML Practitioner, Dataiku DSS Adv Designer, Registered Posts: 1,225 Dataiker
Hi @ecerulm
,
As mentioned in the support ticket on this topic, charts smoothing use a policy that avoids overstepping the real bounds of the data on the Y axis, as this would lead to creating wrongful min and max. One drawback of this policy is that it indeed can create artefacts on the X axis. You need to disable smoothing if you want to ensure no artefacts can occur, neither or X nor Y.
We've taken note of this feature request.
Thanks,