{"id":6119,"date":"2015-08-07T09:00:08","date_gmt":"2015-08-07T13:00:08","guid":{"rendered":"https:\/\/blogs.mathworks.com\/pick\/?p=6119"},"modified":"2015-08-06T08:57:24","modified_gmt":"2015-08-06T12:57:24","slug":"high-dimensional-visualization-with-bubbleplot","status":"publish","type":"post","link":"https:\/\/blogs.mathworks.com\/pick\/2015\/08\/07\/high-dimensional-visualization-with-bubbleplot\/","title":{"rendered":"High-Dimensional Visualization with Bubbleplot"},"content":{"rendered":"

Jiro<\/a>'s pick this week is bubbleplot<\/tt><\/a> by Ameya Deoras<\/a>.<\/p>

Ameya was a fellow MathWorker and also helped us with this Pick of the Week blog. He's now working at a financial firm.<\/p>

I'm always impressed with his MATLAB programs. They are extremely useful and very well written. One of my all-time favorites is Intelligent Dynamic Date Ticks<\/a>, which I also highlighted as a Pick<\/a>. This was a must-have tool, especially before R2014b, when working with data represented as dates. In R2014b, we introduced a new data type datetime<\/tt><\/a>, which makes date handling (including plotting) much easier.<\/p>

Back to this post. bubbleplot<\/tt> is a visualization function for high-dimensional data. If you have ever used scatter<\/tt><\/a> or scatter3<\/tt><\/a>, you probably know that you can already plot four or five dimensional data. With bubbleplot<\/tt>, you'll be able to visualize up to 7 dimensions. Here's an example, borrowing from the examples Ameya provides with his entry.<\/p>

We'll start with a shipping sample data from the Statistics and Machine Learning Toolbox<\/a>.<\/p>

load carsmall<\/span>\r\nOrigin = cellstr(Origin);\r\n\r\n% Create a table for easy overview<\/span>\r\nt = table(Acceleration,Horsepower,MPG,Weight,Model_Year,Cylinders,Origin);\r\nt(1:10,:)\r\n<\/pre>
ans = \r\n    Acceleration    Horsepower    MPG    Weight    Model_Year    Cylinders    Origin\r\n    ____________    __________    ___    ______    __________    _________    ______\r\n      12            130           18     3504      70            8            'USA' \r\n    11.5            165           15     3693      70            8            'USA' \r\n      11            150           18     3436      70            8            'USA' \r\n      12            150           16     3433      70            8            'USA' \r\n    10.5            140           17     3449      70            8            'USA' \r\n      10            198           15     4341      70            8            'USA' \r\n       9            220           14     4354      70            8            'USA' \r\n     8.5            215           14     4312      70            8            'USA' \r\n      10            225           14     4425      70            8            'USA' \r\n     8.5            190           15     3850      70            8            'USA' \r\n<\/pre>
Then create a bubbleplot using these variables. For Color and Shape, the function is able to take categorical data as well. See the help for more information.<\/p>
bubbleplot(t.Acceleration, t.Horsepower, t.MPG,...<\/span>  % X, Y, Z<\/span>\r\n    t.Weight, t.Model_Year, t.Cylinders,...<\/span>         % Size, Color, Shape<\/span>\r\n    Origin,...<\/span>                                      % Text<\/span>\r\n    'FontSize'<\/span>, 6);\r\ngrid on<\/span>\r\nxlabel('Acceleration'<\/span>)\r\nylabel('Horsepower'<\/span>)\r\nzlabel('MPG'<\/span>)\r\n<\/pre>\"\" 
The function is well-written, with good use of inputParser<\/tt><\/a> and plenty of error checking. People wanting to learn good programming practices should have a look. If I were to add one suggestion, I would make it so that the function output (handle objects) gets returned only when it is asked for. When I call the function without an output argument, I don't want \"ans\" created. Just a minor suggestion.<\/p>
Thanks for a cool visualization tool, Ameya. Hope you're doing well!<\/p>
Comments<\/b><\/p>
Give this a try, and let us know what you think here<\/a> or leave a comment<\/a> for Ameya.<\/p>