Demystifying Data Science: Exploring the Top Languages

All programming languages have their advantages and disadvantages. Some of them are slow, and some are fast. Some are low-level languages, and some are not. Some have easy-to-read syntax. Some have libraries that are more suitable to work with statistics, and some are a must to work with tabular data. It doesn’t matter what language you use; libraries will decrease your boilerplate work dramatically. The relatively new and promising languages like Julia compete with the titans of the field like Python and R by offering exciting and useful packages like Pluto.jl, aiming to be the only language you ever need.

Choosing a programming language that fits you can be daunting, especially if this is your first time working with any. This article will discuss the different languages suitable for data science, their strengths, and weaknesses. We will also give you tips on how to pick the right language for your needs. Without further ado, let’s talk about our first and most beloved language out there — Python.

Python, go-to for data science

Let’s start with Python, one of today's most popular programming languages, for data science, machine learning, and almost any other application. Python is a general-purpose language, as you might have guessed already. It’s an open-source language, making it easy for personal and enterprise projects. It also has well-written documentation where you can learn anything about Python anytime. All this builds another advantage — the vast community of Python enthusiasts. And they don’t waste any time. Thousands of libraries that hobbyists and big companies build are shared via PyPI (The Python Package Index with about 450,000 projects), which anyone can download and use for free. It’s the official repository, so most people upload their projects there to reach a wider audience. This is not the only repository of Python packages that you can use. There are also others, such as the Anaconda repository or GIT-like repositories. Pandas, NumPy, scikit-learn, matplotlib, and SciPy are some compelling and valuable libraries you can use with your projects. But before we dive into these, let’s get familiar with Python syntax and its standard library. Let’s start with a basic example.

Suppose that we have a .csv file with Netflix shows named netflix_titles.csv. You can read it by using the built-in functionality.

import csv
from pprint import pprint


with open('netflix_titles.csv', encoding='U8') as f:
  csv_data = list(csv.reader(f))


pprint(csv_data[:2], width=60, compact=True)

It will print the following:

[['show_id', 'type', 'title', 'director', 'cast', 'country',
  'date_added', 'release_year', 'rating', 'duration',
  'listed_in', 'description'],
['s1', 'Movie', 'Dick Johnson Is Dead', 'Kirsten Johnson',
 '', 'United States', 'September 25, 2021', '2020',
 'PG-13', '90 min', 'Documentaries',
 'As her father nears the end of his life, filmmaker '
 'Kirsten Johnson stages his death in inventive and '
 'comical ways to help them both face the inevitable.']]

As you can see, the variable csv_data is a 2D array where the first row contains the names of the columns, and the rest is the actual data. What can we do with this? We can do anything, but implementing it may take much work. Even a simple table analysis can take ages. Wouldn’t it be better to have some functionality out of the box? Yes! Here comes the Pandas package. To install Pandas, write pip install Pandas tabulate in your terminal. It will install Pandas and all dependencies (e.g., NumPy, tabulate). Let’s try to use pandas:

import pandas


df = pandas.read_csv('netflix_titles.csv', encoding='U8')
filtered_df = df[['type', 'title', 'listed_in', 'description']]


print(filtered_df[df['cast'].str.contains('Jenna Ortega',na=False)].to_markdown())

As you can see, reading a .csv is easy. We also filter the columns we want to see in the final result. We are retrieving only the shows with ‘Jenna Ortega’ in the cast. The output is lovely.

|      | type    | title                                        | listed_in                                            | description                                                                                                                                               |
|-----:|:--------|:---------------------------------------------|:-----------------------------------------------------|:----------------------------------------------------------------------------------------------------------------------------------------------------------|
|  530 | TV Show | Richie Rich                                  | Kids' TV, TV Comedies                                | After turning his veggies into green energy, Richie Rich earns a trillion dollars overnight and a new life filled with nonstop fun and adventure.         |
|  855 | TV Show | Jurassic World Camp Cretaceous               | Kids' TV, TV Action & Adventure, TV Sci-Fi & Fantasy | Six teens invited to attend a state-of-the-art adventure camp on Isla Nublar must band together to survive when the dinosaurs break out of captivity.     |
| 1218 | Movie   | YES DAY                                      | Children & Family Movies, Comedies                   | A mom and dad who usually say no decide to say yes to their kids' wildest requests - with a few ground rules - on a whirlwind day of fun and adventure.   |
| 2029 | Movie   | The Babysitter: Killer Queen                 | Comedies, Horror Movies                              | Two years after Cole survived a satanic blood cult, he's living another nightmare: high school. And the demons from his past? Still making his life hell. |
| 3324 | Movie   | The Massively Mixed-Up Middle School Mystery | Children & Family Movies, Comedies                   | After the school principal gets pranked, a curious crew of preteen super sleuths tests their detective skills to solve an underwater ruse.

As you can see, working with a 3rd party library is simple, and you don’t need to think about lots of mind-boggling stuff. For example, this output is properly formatted in a table-suited markdown syntax!

Let’s look at some of the most used libraries:

• NumPy — the library that lets you work with huge arrays of the same type efficiently. It’s written in one of the low-level programming languages (C) and is well-optimized! It has many mathematical functions, random generators, linear algebra routines, Fourier transforms, and more. NumPy is the backbone of most scientific data science and machine learning libraries.
• Pandas — the library to work with all kinds of tabular data. It’s not a coincidence that we touched it during our example. Along with all that, filtering, grouping, reading, and writing pandas is good with dates! Series is the second most used class in pandas.
• Matplotlib — the library to create all types of visualization. If you need animated, interactive, or static plots, matplotlib is the way to go.
• SciPy — the library that provides algorithms for optimization, integration, interpolation, eigenvalue problems, algebraic equations, differential equations, statistics, and many other classes of problems.
• PyTorch — the library for deep learning models that supports computer vision, NLP, and more
• TensorFlow — if you are interested in machine learning applications, this library is for you
• scikit-learn — machine learning, data mining. It has a bunch of analytics algorithms for things like finding patterns, making predictions, grouping things, reducing complexity, picking features, and more.

Let’s summarize:

• Python is easy to learn and use. The syntax is clear and readable.
• Python uses a lot of memory and may require much additional work to reduce it.
• Python is popular in machine learning and deep learning, with libraries such as TensorFlow, PyTorch, and scikit-learn.
• Python has powerful data manipulation capabilities, thanks to libraries such as NumPy and Pandas.
• Python is an interpreted language, meaning it is slower than compiled languages.
• Python has a huge and active community of developers and data scientists who provide many libraries for data analysis.
• Python is a general-purpose language that can be used for various tasks.
• Python can be used within a Jupyter Notebook. They can be run locally or by using Google Colab. This fact makes your work much simpler to share, collaborate with others and test everything on the fly!

Julia, respected by scientific community

Julia is the next language on our list, explicitly created for scientific and numerical computing. It is a versatile and compiled language that doesn't require extra layers to work with arrays, making it fast and efficient. It can achieve performance levels comparable to those of C! Additionally, Julia is significantly faster than Python, although it's worth noting that Python uses C libraries to keep up. Julia's syntax is straightforward, expressive, and more user-friendly than Python's, which makes it an excellent choice for beginners who want to focus solely on science.
Like Python, Julia is a general-purpose language for scientific computing, numerical optimization, machine learning, data science, visualizations, etc. In addition, a growing and enthusiastic community also works on creating custom and helpful libraries. Julia has several extensive libraries for machine learning and data science. Besides of ScikitLearn.jl and TensorFlow.jl, it’s including Flux.jl for neural networks, DataFrames.jl for tabular data, MLJ.jl for building and evaluating machine learning models, and Pluto.jl as a reactive notebook! Let's put some code into practice to see if it’s worth it.

# first, let's load some packages.
using RDatasets
using DataFrames
using Plots


# let's load our dataset next
iris = dataset("datasets", "iris")


# finally, let's make a plot!
scatter(iris.PetalLength, iris.PetalWidth,
       group = iris.Species, 
       xlabel = "Petal Length",
       ylabel = "Petal Width")

The output is:

As you can see, this is a classic iris dataset plotted using Julia within five lines. Another point for Julia!

Another exciting feature of the language is the multiple dispatch, which is the ability to pick the correct function based on the provided arguments.

function foobar(x::Array)
  println("foobar(x::Array)")
end


function foobar(x::Tuple)
  println("foobar(x::Array)")
end


function foobar(x::Tuple, y::Int)
  println("foobar(x::Tuple, y::Int)")
end


function foobar(x::Tuple, y::Float64)
  println("foobar(x::Tuple, y::Double)")
end


foobar((1, 2)) # foobar(x::Array)
foobar([1, 2]) # foobar(x::Array)
foobar([1, 2], 5) # MethodError
foobar((1, 2), 2.0) # foobar(x::Tuple, y::Double)
foobar((1, 2), 0) # foobar(x::Tuple, y::Int)

Multiple dispatch can handle different kinds of data and operations, which in the end, will make your code cleaner and simple to work with.

Let’s summarize:

• Julia is a fast and efficient language that can handle large-scale data processing and machine-learning tasks.
• Julia is a general-purpose and multi-paradigm language that allows you to write code in any way you want: OOP, functional, or declarative.
• Julia has a multiple dispatch system which makes your code more efficient and easy to maintain.
• Julia has a growing and enthusiastic community with a constantly growing number of high-quality libraries.
• A reproducible environment makes sharing and testing your code and collaborating with others as easy as using Python.
• There is little to say about the cons for Julia. It’s a fast-growing language that can fix any of these in no time!

‍R, best for statistical modeling

R is a system used for statistical analysis, computation, and graphics. It includes a language, a run-time environment with graphics, a debugger, access to specific system functions, and the ability to run programs stored in script files. It is important to note that R is not a general-purpose programming language but is explicitly designed for statistical computations and analysis. Let's dive into it further.

R is a data manipulation language. Like the two other languages we’ve discussed, you can read, clear, and transform data from various file formats, such as CSV, JSON, XML, SQL, and so on. You can use R to perform any kind of EDA, such as summary statistics, finding outliers, and handling missing values. Machine learning tasks are also possible—regression, classification, clustering, and natural language processing. You can also evaluate and compare different statistical models, tune, and deploy your models. Statistical computing and analysis, such as hypothesis testing, ANOVA, linear models, time series analysis, and Bayesian inference, are also possible with R.

You can also use R to simulate data and perform Monte Carlo methods. Data visualization is one of the strongest sides of the R language. You can create charts, graphs, and many other interesting and unusual visualization types by using libraries such as ggplot2, plotly, shiny, or without any! Let’s find out if we can plot anything without 3rd party libraries.

x <- seq(-pi, pi, 0.1) 
plot(x, cos(x),
   pch = c(4, 5, 6, 8, 15),
   col = c("red", "blue", "violet", "green"),
   xlab = expression(x),
   ylab = expression(cos(x)))

And the output is:

You read that right! You can plot without importing any libraries!

Let’s summarize and find out if it has some cons too:

• R was designed as a statistical programming language. Created by statisticians for statisticians with lots of built-ins for data analysis, modeling, testing, and visualization.
• R has a big community of developers creating libraries for data science tasks and machine learning.
• R is a good choice for data science if you are interested in statistical operations or want to create interactive dashboards or reports.
• However, it has a steep learning curve, and because of its intuitive syntax (many ways to write the same thing and no preferable coding style), it may create even more problems at the beginning.
• Like Python, it’s slow and doesn’t handle memory well.
• The lack of built-in security tools may cause some troubles with sensitive data.
• The lack of parallel computing puts us even lower in our list of languages.