Popularity of Python, R, and Matlab

Matlab has been used for years. In recent years, with the rise of linux, open source community projects such as python and GNU R have found increasing use and a recent article in NY times wrote about the rise of R. I took the time to create graphs to compare trends in popularity of three algebra computing platforms. This could help in predicting which is the future platform for scientific computing.

People use different software packages for data analysis including excel, SAS, SPSS, or perl. I also have a post about C++ software for number crunching. I could have taken all these alternatives into account, however, I think desiderata for a language should be these three:

• It should be possible to do fast-prototyping. This includes:
  
  • It should be a scripting language.
    
  • It should have a lot of available libraries
• It should be cross-platform (at least linux and windows) to be portable.
  
• It should be fast.

Python (with SciPy and NumPy), R, and matlab meet these desiderata. All of them are cross-platform and scripting languages. All of them are reasonably fast and allow the easy integration of C, C++, and Fortran code to achieve more speed-up. There are a lot of libraries available especially for R and matlab (I am not knowledgeable about python libraries).

So, in this post I want to look at the popularity of python, R, and matlab. I first look at the numbers of citations in scientific publications, then I link to google trends to see visits to project websites.

To look at the scientific zeitgeist, I compiled numbers of citations from scholarly articles of the three platforms counting them in citeseerX and google scholar.

Matlab is the oldest software package. Matlab v. 1.0 was released 1984. Guido van Rossum published version 0.9.0 of python in 1991. The R mailing list started 1997; this was with version v. 0.16.

Google scholar has a huge index of scholarly articles and searches within the text of the articles (which is not done by all search engines). CiteseerX has much fewer articles and article count differs hugely over years with much fewer articles in the last three years. Therefore the curves were normalized by the total article count for each year.

I am conscious that searching "python" many hits are not relevant, however I think the graphs here nonetheless can give some general idea to python's relative importance. Results should be taken with a grain of salt.

Google scholar returns different numbers of hits and different hits over several trials, a fact which I ignored for simplicity. There were also inconsistencies in the "recent articles" feature. You can say you want articles published earliest in some year and you would expect the more years back you go, the more articles you find. However this is not the case always, so the google scholar graph is restricted to 2001 onwards.

Data as of January 8th, 2009.


Here the graphs (click to increase):

CiteSeerX

Google scholar

Matlab comes out as the big winner. Python and R find increasing popularity. From both graphs you can see that R and python are on the rise. CiteseerX does not have as many citations as google scholar (between 1,500 and 30,000 for each year between 1990 and 2008). Because of big fluctuations in the matlab curve in the google scholar count, I do not dare to make any conclusion about matlab.

Alexa and google trends among other services alow you the comparison of websites. Here is the link to google trends' comparison between R, matlab, and python. Surprisingly, python and matlab are not that different in page visits.

--
Graphs created in R. Thanks to Statsmethods.net for the explanation on the plot and points functions in R.