Introductory videos for Explanatory Model Analysis with R

April 3, 2020
By

[This article was first published on Stories by Przemyslaw Biecek on Medium, and kindly contributed to R-bloggers]. (You can report issue about the content on this page here)
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.

Remote teaching at my university encouraged me to prepare some video materials for Explanatory Model Analysis techniques, i.e. techniques of exploration, explanation and visualisation of predictive models.

The pyramid for Explanatory Model Analysis. Left part is focused on a single observation (instance-level or local explanations). Right part is focused on the whole data set (called dataset level or global explanations). In the first row are raw model predictions and model statistics. The second row is related to the analysis of model parts. The third row describes the profile of model behaviour depending on the values of individual variables. The fourth row corresponds to the residual diagnostics.

For example, for a model that estimates odds of default in credit scoring, the model level analysis focuses on explaining the model’s behaviour for a selected population of customers.

The ebook describing these techniques was already available at https://pbiecek.github.io/ema/.
Today I have published a youtube playlist with short videos showing how to use these techniques in R.

For now, the first 5 videos are available for instance level analysis. These videos are focused on the DALEX package for R. You can read about other R packages for XAI and more about the methodology in the ebook.

Here are these videos

  1. Gentle introduction to the topic of exploration, why this topic is important and how it fits into the model building process.

2. Introduction to the DALEX package and other tools from the DrWhy.AI family (e.g. modelDown and modelStudio)

3. Instance level attribution with break-down method for determining variables important for a prediction. The break-down method decomposes predictions of any model into parts that can be attributed to model variables.

4. Instance level attribution with Shapley values and break-down with interactions. Both extends break downs possibilities.

5. Presentation of the ceteris paribus method for the analysis of model response profile for selected variables for the indicated observation.

The whole playlist is at https://www.youtube.com/playlist?list=PLGzKiXahhU63NmyM7sALiVtFYTUYEpTVm

Supplementary materials are at http://dalex.drwhy.ai/.

Now we’re working on videos for the python version of this package. Stay tuned!

Questions or comments? Let me know at ema at drwhy.ai.

Warm thanks to everyone who helped me to prepare these videos, especially Wojciech Kretowicz, Huber Baniecki, Alicja Gosiewska, Anna Kozak and Katarzyna Woźnica.

To leave a comment for the author, please follow the link and comment on their blog: Stories by Przemyslaw Biecek on Medium.

R-bloggers.com offers daily e-mail updates about R news and tutorials about learning R and many other topics. Click here if you're looking to post or find an R/data-science job.
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.



If you got this far, why not subscribe for updates from the site? Choose your flavor: e-mail, twitter, RSS, or facebook...

Comments are closed.

Search R-bloggers

Sponsors

Never miss an update!
Subscribe to R-bloggers to receive
e-mails with the latest R posts.
(You will not see this message again.)

Click here to close (This popup will not appear again)