Research Data Visualization Tools to build project

[RaghavSaini01]

As developers, we want to take some time to scope out and use the most effective data visualization library, framework, or tool for the purposes of our report.

This part involves researching different tools in the following ways:

• Read about the library: what language does it work with, is it compatible with your OS or not, is it paid, etc
• Watch a tutorial: See the code writing in action, what basic setup looks like, and how functions and other parts of the library are used
• Run sample code/write your own small bits: No better way to get good at a library than by seeing it in action! Bigger libraries will usually give you sample code you can run officially from the library devs themselves and for others you may be able to find 3rd party github repos. Running them to see what output looks like and what they're capable of is great. Even better would be if you made up some random data and just tried stuff with it on your own!

[RaghavSaini01]

Probably the best one-off starting point I can find: https://medium.com/sciforce/best-libraries-and-platforms-for-data-visualization-b986a43aee3f

My personal recommendation would be to go between one of R or Python, but Js could be worth exploring if y'all wanna learn too :D

[ayanapatidar]

One of the first steps we'd probably want to take is figure out whether we want our MVP to be interactive or not. If interaction isn't something we think of as particularly important, my vote is to go with Python. The language is optimized for data work, and there's a couple of tools I found that would be great to have at our disposal:

-Seaborn (really nice and clean visualization tool library built off of matplotlib) -This set of tutorials on doing word analysis in Python using libraries like numpy, pandas, sklearn, nltk, and the aforementioned seaborn. The link leads to one specific tutorial, but this user has a set of articles that all deal with this specific type of analysis, if we choose to pursue the word analysis route. The tutorials are paywalled after a certain amount, but you can get around it by using incognito mode.

If we do want something that's interactive though, our best bet will probably be d3.js. Here's some documentation and tutorials:

-Official Documentation. This comes with some examples that can be pretty useful to plug actual data into. -Technology Fundamentals for using d3.js. Another source of the same basic idea is here. -Video Tutorials -Data Binding -Cheat Sheet! -We'd then need to set up a server and a web application to be able to host the site.

This isn't to say that we couldn't make a Python program interactive, we definitely could. In visualization terms however, d3.js will definitely be a lot cleaner, more accessible, and more interactive. I think we might have to take a vote on which direction we want to move in! ^_^

[neha-vard]

This is a good source I found that includes what each library interacts well with: https://www.monterail.com/blog/javascript-libraries-data-visualization

We may want to consider using a library that works well with React if we want to integrate it easily into a website. One such library I found is Frappe Charts. Though there are many others, this one stood out because it includes maps, which may be helpful for our visualizations.

However, d3.js, which Ayana mentioned is very vast and we can do a lot with it. Though it clashes a little with React, there are workarounds.

[RaghavSaini01]

https://sites.northwestern.edu/researchcomputing/2022/02/03/what-is-the-best-interactive-plotting-package-in-python/ A couple of them are “mash-up libraries” between Python and js that seem to be really neat and versatile, and Bokeh, Plotly, and mpld3 especially seem to have good documentation and be really good for more exploratory, interactive type data vizes rather than an explanatory one

--Saket

[michelleruu]

Firstly, I do think that something interactive would be fun to have (@ayana), and I think I'm currently leaning a little more towards something like d3.js just because it seems like it would be more clean/easier.

I also think what neha-vard posted had a lot of really great resources. I think that the Frappe Charts would be a good choice, and I was looking into React-vis and it seems like a decent choice as well. Honestly I'm down for anything, the "mash-up libraries" that Raghav posted also seem very clean and not too hard to work with.