Machine learning and data analysis are two areas where open source has become almost the de facto license for innovative new tools. Both the Python and R languages have developed robust ecosystems of open source tools and libraries that help data scientists of any skill level more easily perform analytical work.
The distinction between machine learning and data analysis is a bit fluid, but the main idea is that machine learning prioritizes predictive accuracy over model interpretability, while data analysis emphasizes interpretability and statistical inference. Python, being more concerned with predictive accuracy, has developed a positive reputation in machine learning. R, as a language for statistical inference, has made its name in data analysis.
That isn't to pigeonhole either language into one category—Python can be used effectively as a data analysis tool, and R has enough flexibility to do some good work in machine learning. There is a multitude of packages for both languages that seek to replicate the functionality of the other. Python has libraries to boost its capacity for statistical inference and R has packages to improve its predictive accuracy.
Python's machine learning and data analysis packages
Even though Python is naturally disposed toward machine learning, it has packages that further optimize this attribute. PyBrain is a modular machine learning library that offers powerful algorithms for machine learning tasks. The algorithms are intuitive and flexible, but the library also has a variety of environments to test and compare your machine learning algorithms.
Scikit-learn is the most popular machine learning library for Python. Built on NumPy and SciPy, scikit-learn offers tools for data mining and analysis that bolster Python's already-superlative machine learning usability. NumPy and SciPy impress on their own. They are the core of data analysis in Python and any serious data analyst is likely using them raw, without higher-level packages on top, but scikit-learn pulls them together in a machine learning library with a lower barrier to entry.
When it comes to data analysis, Python receives a welcome boost from several different packages. Pandas, one of its most well-known data analysis packages, gives Python high-performance structures and data analysis tools. As is the case with many of Python's packages, it shortens the time between starting a project and doing meaningful work within that project. If you really want to stick with Python and get as much R functionality as you can, RPy2 offers all of R's major functionality. This gives you the best of R in Python natively.
R's machine learning and data analysis packages
R, like Python, has plenty of packages to boost its performance. When it comes to approaching parity with Python in machine learning, Nnet improves R by supplying the ability to easily model neural networks. Caret is another package that bolsters R's machine learning capabilities, in this case by offering a set of functions that increase the efficiency of predictive model creation.
But data analysis is R's domain, and there are packages to improve it beyond its already-stellar capabilities. Packages for the pre-modeling, modeling, and post-modeling stages of data analysis are available. These packages are directed at specific tasks like data visualization, continuous regression, and model validation. With all of these cross-functional libraries and packages, which language should you drag into the data battlefield with you?
Python for machine learning and data analysis
If you have some programming experience, Python might be the language for you. Python's syntax is more similar to other languages than R's syntax is. Python's readability is also nearly unmatched, as it reads much like a verbal language. This readability emphasizes development productivity, while R's non-standard code could lead to stutters in the programming process.
Python is well known as a flexible language, so if you plan to move on to projects in other fields when your machine learning or data analysis project is done, it might be a good idea to stick with Python so you aren't required to learn a new language.
Python's flexibility makes it a great choice for production use because, when the data analysis tasks need to be integrated with Web applications, for example, you can continue to use Python instead of integrating with another language. R is a great data analysis tool, but is fairly limited in terms of what it can accomplish beyond data analysis.
If you're completely new to programming, and therefore unfamiliar with “standard” syntax, the learning curve for both languages is roughly the same. However, if the goal is to push past the basics of machine learning and data analysis, Python is probably a better choice. This is especially true considering the addition of scikit-learn to Python's arsenal of packages. The package is well maintained and actively in development. R might have a greater diversity of packages, but it also has more fragmentation and less consistency across those packages.
R for machine learning and data analysis
To date, R has primarily been used in academics and research. This is beginning to change, though, as R usage expands into the enterprise market. R was written by statisticians and it shows—basic data management tasks are very easy. Labeling data, filling missing values, and filtering are all simple and intuitive in R, which emphasizes user-friendly data analysis, statistics, and graphical models.
Since R was built as a statistical language, it has great statistical support overall. It represents the way statisticians think pretty well, so for anyone with a formal stats background it feels natural. Packages like statsmodels provide solid coverage for statistical models in Python, but the ecosystem of statistical model packages for R is much more robust. As far as beginner programmers are concerned, R makes exploratory work easier than Python because statistical models can be written with just a few lines of code.
R's closest answer to pandas is probably dplyr, but it is more limited than pandas. That might sound negative, but dplyr has the benefit of being more focused, which makes discovering how to perform a task much easier. Dplyr is also more readable than pandas.
Choosing your language
The main issue with R is its consistency. Algorithms are provided by third parties, which makes them comparatively inconsistent. The resulting decrease in development speed comes from having to learn new ways to model data and make predictions with each new algorithm you use. Every package requires a new understanding. Inconsistency is true of the documentation as well, as R's documentation is almost always incomplete.
However, if you find yourself in an academic setting and need a tool for data analysis, it's hard to argue with choosing R for the task. For professional use, Python makes more sense. Python is widely used throughout the industry and, while R is becoming more popular, Python is the language more likely to enable easy collaboration. Python's reach makes it easy to recommend not only as a general purpose and machine learning language, but with its substantial R-like packages, as a data analysis tool, as well.
If you don't already know R, learn Python and use RPy2 to access R's functionality. You'll be getting the power of two languages in one, and Python is production-ready because most companies have production systems ready for Python. This isn't true for R. Once you learn RPy2, the jump to pure R isn't very daunting, but moving in the opposite direction is considerably more difficult.
Both Python and R have great packages to maintain some kind of parity with the other, regardless of the problem you're trying to solve. There are so many distributions, modules, IDEs, and algorithms for each that you really can't go wrong with either. But if you're looking for a flexible, extensible, multi-purpose programming language that also excels in both machine learning and data analysis, Python is the clear choice.