Top Python Data Science Libraries

Data science with Python has really started to bloom in recent years with Python becoming the most popular environment for data scientists. Python's biggest power for data science lies in the great Python libraries and modules available for data science. In this article I am going to introduce you to different Python libraries for working with data science. I am not going to talk about all the libraries available there, but the ones that are most popular and commonly used.

NumPy

NumPy is the most important Python library for scientific computing with Python. This is a dependency of many other scientific Python libraries. NumPy provides a lot of convenient functions and data structures to work with numbers, linear algebras, random number generation, Fourier transformation, etc.

Python is inherently slow for its dynamic behavior. So, a library written in Python could be slower than a library in C/C++. But, scientific computation and working with large arrays and multidimensional arrays requires the system to be faster. Python is great for working without numbers without thinking about limitations or the size of the number. Again, Fortran has a reputation of working with numbers. To make everything together and still keeping it faster NumPy was created as a C extension to Python and with the help of Cython.

Matplotlib

Matplotlib is a Python library for data visualization. Data is worthless when it cannot tell a story visually. It can visualize structured data in multiple forms and with a lot of customization. It can also output the visualization in multiple file types like, JGP, PNG, PMP, GIF, SVG, PDF, etc. You can quickly and easily make line graphs, pie charts, scatter plots, histograms, and other types of figures. Matplotlib was primarily created for 2D graphics, but it is also possible to create 3D graphics and effects.

Matplotlib was created by John D. Hunter who faced a lot of difficulties working with Matlab. Matlab does not provide a rich programming environment like Python, it is not flexible either, and again it is not open source like Python. All the limitations resulted into the creation of Matplotlib. Now, when there is a need for data visualization in Python we first remember the name of matplotlib.

NLTK

NLTK or Natural Language Toolkit is a collection of libraries and tools for natural language processing in Python. It works with human language data. NLTK provides facilities for text parsing, tagging, tokenization, semantic reasoning, classification, etc. It comes with a large number corpora and lexical resources for various human languages.

Pandas

Panda is a Python library that provides various high level convenient data structures, functions, and classes for easy and fast data analysis and manipulation operations. It is built on top of NumPy and thus it provides interchangeability between many other popular Python libraries that uses NumPy. Use of NumPy data structures and functions makes it relatively faster. It also works with a wide range of data format.

SciPy

SciPy is a collection of modules that provide optimization, linear algebra, integration, interpolation, signal processing, image processing, etc. SciPy is also built on top of the versatile library NumPy and thus reuses all of its convenient features.

Scikit-learn

Scikit-learn is a machine learning library for the Python programming language. It provides very commonly used machine learning algorithms with a consistent API. It can be used to implement commonly used algorithms on various data sets. Some features include: classification, regression, clustering, etc.

Theano

Theano is similar to NumPy. It provides various numerical computation for Python. It can run both on CPU and GPU seamlessly. It helps define, optimize and evaluate mathematical expressions. Theano expresses computations using NumPy-esque syntax.

Scrapy

Scrapy is not a data analysis Python library, instead it is a data aggregation script development library built on top of Twisted. In data science we cannot work with data unless we get it for various sources. The web for example is a good resource for data. But data is not available through some clean API through these websites. So, we need to scrape them and extract the data out to put in a certain format. It is a hard task to create a scraper from scratch. With the help of Scrapy we can use it as a framework to scrape the web.

Beautiful Soup

On the web data is not formatted in most cases. On web pages it is easy for human eyes to understand data quickly, but for machines it is a nightmare unless we make the data suitable for the machine. A web page is a dirty pile of special tags and texts. We need to parse it and extract the data we want. Beautiful Soup is the best library to parse a malformed web pages. It can be used with Scrapy to parse the scraped pages.

RE

Re or RegEx or Regular Expression library of Python is a built in Python library. In the crowd of a lot of third party modules people forget this great library built-in with Python distribution. re is a Perl compatible regular expression library that can work with regular expressions and find, replace or manipulate text or binary data. re can be used with both binary strings and unicode data.

We often need to search through a lot of text or binary data in our data analysis projects. With common string finding or manipulation it is a nightmare to carry out a simple task. If we find some pattern in texts or bytes we can use the re library to deal with them.

Conclusion

Python has a lot more libraries and tools available than described in this article. In this article I have described the most used ones and at the end I provided a built in library that most of the people forget that it can be used like a survival weapon. Check back soon for more articles and tutorials on different Python libraries including libraries related to data science, machine learning and artificial intelligence.