Prolog and data science

[jan]

I got a link to a blog from @emiruz in my linked-in profile. This seems worth sharing here, so I asked him whether that was ok and got these two replies:

@emiruz

Yes of course, that’s kind of you. I’m a data scientist by profession, I started learning Prolog in April 2023, and I keep the blog as a way of committing to my experiments, so since then I’ve written a whole bunch about Prolog. My Prolog is getting better every week – it wasn’t very good when I started

Here is a complete list of articles:

https://emiruz.com/post/2023-11-13-domicles/
https://emiruz.com/post/2023-10-18-minimal-prob-prolog/
https://emiruz.com/post/2023-10-15-logical-data-analysis/
https://emiruz.com/post/2023-10-06-dominoes/
https://emiruz.com/post/2023-07-30-optimal-stopping/
https://emiruz.com/post/2023-05-12-semi-supervised-clustering/
https://emiruz.com/post/2023-04-30-prolog-for-data-science/

@emiruz

PS: purely from a data science perspective, this article was important to me:

Better data analysis with logic programming - Emir's blog

It is where I realised for the first time that Prolog offers me a completely different perspective: I should not use it as yet another language to run a regression or some machine learning library from, I should instead analyse the data relationally: logic first.

The Prolog model in the article is simple and more useful than most other treatments. There is a “Why is this better?” section

I’m surely going to read more of that. The initial link was Prolog for data science - Emir's blog. This was something new to me. I used Prolog for data science with a PhD student (Tessel Boogaard). There we used SWISH and the embedded R connection. The task of Prolog was to preprocess the raw data, preparing a dataset that was ready for statistical analysis. In the article above, Prolog is used for post-processing, i.e., working on the output of the numerical analysis.