added k-means clustering in php to eda page

[jacobemerick]

After learning about k-means clustering I decided to try to recreate it in PHP. It's not perfect, but it works, and is surprisingly fast. I was able to use the application to estimate clustering on Google Maps geographical data (blog: http://blog.jacobemerick.com/web-development/k-means-clustering-on-hiking-adventures/).

[seankross]

This is a great blog post! Unfortunately I'm unable to merge it for the following reasons:

• There's no discussion about how k-means clustering works or why one would want to use k-means for this kind of data.
• There's no discussion about your implementation of k-means, and why it might be better/faster than other PHP implementations.

Before sending a pull request to this site you should ask yourself:

1. Does my contribution teach?
2. Does the content of my contribution clearly address topics in the Data Science Specialization?
3. Could my contribution be seamlessly integrated into the canonical course materials?

[jacobemerick]

It's all good - the blog post is not in the pull request. The blog post was just meant to illustrate the k-means clustering tool and is not linked. The tool is the PR, not the post: applying a topic of the class (k-means clustering) and delving into the logic underneath it to construct it in another language. You're right about the post, it isn't really that instructive, but the PR links to the repo where the tool is.

If you don't think the tool is relevant enough (it is in another language, after all) I'll remove the PR.

Thanks sir!

[seankross]

This is certainly fertile ground for a relevant contribution. If you went into detail in a post describing how your code works (and therefore how k-means works), your reference for this particular k-means algorithm, and the challenges associated with porting it to PHP I'd be more willing to merge. This site is meant for sharing educational content developed by the community, not standalone pieces of software.

Still - we want potential contributors to err on the side of sending a pull request when in doubt. Anyway it's a nice library!

[jacobemerick]

Gotcha. If I end up writing up an analysis of the tool and the different elements addressed in a separate post I'll be sure to send another PR. Thanks for the explanation, sir!