Tuesday, May 10, 2016

Exploring Data Visualization

I spent this quarter looking at various ways to make data analysis more complex, going past pie charts and bar graphs, but also more accessible to the average user, by presenting and organizing it in a more visually intuitive way. The two main data sets I used for this exploration are: 1) A personal, mid-size data set. Every day in 2015, I rated how I felt about that day from 0-100 (though my numbers ended up being 40-70). 2) A very large set of survey answers collected from 15,000+ high school students across the nation, including questions on health, video games and television, online behavior, and bullying. Using these two data sets, I examined the usefulness of 5 different, freely accessible data analysis services: Plotly Free, Tableau Public, Google Sheets/Charts, Datacopia, and IBM Watson.

Plotly Free (3/10)

Ease of Use: Convoluted (2/10)
Learning Curve: Intuitive (7/10) when graphing, Steep (4/10) when navigating around the site
Visuals: Decent (7/10)
Customization and Versatility: Decent (6/10)
Privacy: Allows 2 private projects before requiring public publishing (4/10)
Key Features and Recommended Usage: Accessible and editable online, open-source, programmer APIs, inexpensive pro version

Plotly seems like a good all-purpose candidate in theory—and, to be fair, some friends have had good experiences with Plotly. However, I spent most of my time with Plotly waiting on endless loading screens, wondering if something was loading or not when I clicked something, and attempting to figure out the mechanics and layout of things. About the privacy score: while I don't mind having a limit on private projects in the free version of a service, I don't appreciate having pop-ups when I exceed the limit, imploring me to upgrade to Plotly Pro. When I was actually organizing data, I found Plotly to be almost pleasant. But if my experience of frustration would be similar to anyone else's, I
cannot recommend using Plotly.

Tableau Public (9/10)

Ease of Use: Smooth (9/10)
Learning Curve: Somewhat difficult (5/10)
Visuals: Excellent (9/10)
Customization and Versatility: Good (8/10)
Privacy: All projects are saved to a public account (6/10)
Key Features and Recommended Usage: Program/application, box-and-whisker plots (I cannot express how much I love Tableau's box-and-whisker plots)

I spent most of my time with Tableau Public, and if you really want to explore a variety of analysis perspectives, Tableau is good for the job; it offers many chart types in a selection box that are choosable once the data is in. Because Tableau won't just analyze data on its own, only graph, you need to know what you want and what you're doing with the data. The interface is drag-and-drop into a multitude of work areas, which means it's easy to look at, but some of it gets lost in "what does this do?" It is all explained in Tableau's comprehensive how-to videos, but many casual users won't have the time to watch them all.



You can find the rest of the charts that I created with Tableau here.

Google Sheets/Charts (7/10)

Ease of Use: Smooth (9/10)
Learning Curve: Simple (9/10)
Visuals: Basic (5/10)
Customization and Versatility: Basic (5/10)
Privacy: Depends on your opinion of Google. For me, Trusted (8/10)
Key Features and Recommended Usage: Online cloud access, live data, easily shared

The chart and graph maker in Google Sheets is basic but definitely intuitive to use. Most everyone I know has had some experience with it; simply put your data into a spreadsheet, select it, and hit Graph. The little widget that comes up also offers some likely graphs that you might want. I find that Google Sheets gets the job done, but is very limited in aesthetic customization.


Datacopia (6/10)

Ease of Use: Simple (8/10)
Learning Curve: Simple (10/10)
Visuals: Basic, Colorful (6/10)
Customization and Versatility: None (1/10)
Privacy: Private (10/10)
Key Features and Recommended Usage: Simple

Datacopia is even easier to use than Google Sheets, a feat I find impressive in itself. However, it sacrifices customization for that ease. Once you upload your data, Datacopia generates a large (and I mean large. 100 or so) set of premade charts and you can pick one. The graphics are appealing, but if what you want isn't on there, you're out of luck. I generally would not recommend this for all but the simplest projects. One good note is that it promises that the data you upload does not leave your computer—completely private, if that's a must-have for you.


Watson IBM (7/10)

Ease of Use: Decent (6/10)
Learning Curve: Difficult (5/10)
Visuals: Decent, lacking in customization (6/10)
Customization and Versatility: Fair (5/10)
Privacy: Private Cloud Account (9/10)
Key Features and Recommended Usage: Handles big data well, business-oriented analytics, offered data insights

Watson IBM Data Analytics looks like a fantastic service for those doing heavy lifting with their data. It offers a variety of chart types. The best thing about Watson is probably its Exploration—an AI screens your data and suggests questions that might be answered by correlations it finds. Watson can detect gaps in your data—missing information that may compromise the quality. It can also take the correlations it finds and attempt to detect the why: which factors affect each other. Combined, it makes a powerful tool. However, I found it lacking in aesthetic customization, being unable to change the individual color, only pick from preset palettes. For casual users, it may also be hard to navigate the interface.



1 comment:

  1. Informative post . Like and appreciate your helpful post and thanks . Clipping Path

    ReplyDelete