Do Data Scientists Code?

In a word, yes. Data Scientists code. That is, most Data Scientists have to know how to code, even if it’s not a daily task. As the oft-repeated saying goes, “A Data Scientist is someone who’s better at statistics than any Software Engineer, and better at software engineering than any Statistician.”

The amount of programming (a.k.a. coding) they actually do, however, depends on their role and the tools they’re using. A few examples of the things Data Scientists can expect to program:

• Analysis scripts, usually in R or Python, with the intention of generating actionable insights.
• Prototypes of digital products. Using Python, the goal is generally to prove the efficacy of a new product or feature, which allows a Developer to then build it.
• Production code. In smaller companies, Data Scientists often have full responsibility for this, and may have to make use of Ruby on Rails or Java (in addition to the more commonly used data science languages) to achieve this.

What Programming Languages Do Data Scientists Use?

Python, R, SQL, and Java are some of the most popular programming languages Data Scientists use. Let’s take a closer look at how Data Scientists use these programming languages and more.

Python

With a manageable learning curve and an array of libraries that allow for nearly endless applications, Python is the top programming language of choice for the many Data Scientists who appreciate its accessibility, ease of use, and general-purpose versatility. In fact, BrainStation’s 2019 Digital Skills Survey found that Python was the most frequently used tool for Data Scientists overall.

Since its introduction in 1991, Python has built up an ever-growing number of libraries dedicated to carrying out common tasks, including data preprocessing, analysis, predictions, visualization, and preservation. Meanwhile, Python libraries like Tensorflow, Pandas, and Scikit-learn allow for more advanced machine learning or deep learning applications. Asked about their preference for Python over R, Data Scientists cited Python’s tendency to be faster than R, and better for data manipulation.

R

A free, open-source programming language that was released in 1995 as a descendant of the S programming language, R offers a top-notch range of quality domain-specific packages to meet nearly every statistical and data visualization application a Data Scientist might need—including neural networks, nonlinear regression, advanced plotting, and much more. Its visualization library ggplot2 is a powerful tool, and R’s static graphics can make it easier to produce graphs and mathematical symbols and formulae.

Yes, Python does have a speed advantage over R (and R does feature a steeper learning curve than the more approachable Python), but for specific statistical and data analysis purposes, R’s vast range of tailor-made packages gives it a slight edge. It’s worth noting that, unlike Python, R isn’t a general-purpose programming language—it’s intended to be used specifically for statistical analysis.

SQL

SQL, or “Structured Query Language,” has been at the core of storing and retrieving data for decades. SQL is a domain-specific language used for managing data in relational databases—and it’s a must-have skill for Data Scientists, who rely on SQL to update, query, edit, and manipulate databases and extract data. Though SQL is not as useful as an analytical tool, it’s highly efficient and crucial for data retrieval. This makes SQL a particularly helpful tool for managing structured data, especially within large databases. Since SQL is a core skill, it’s fortunate that its declarative language is quite readable and intuitive.

Java

As one of the oldest general-purpose languages used by Data Scientists, Java owes its usefulness, at least in part, to its popularity: many companies, especially big, international companies, used Java to create backend systems and applications for desktop, mobile, or web. Skill with Java is increasingly attractive thanks to Java’s ability to weave data science production code directly into an existing database. It’s also highly regarded for its performance, type safety, and portability between platforms. It bears mentioning that (really) big data computation application Hadoop runs on the Java virtual machine (JVM)—yet another reason Java is a must-have skill for Data Scientists.

Scala

User-friendly and flexible, Scala is the ideal programming language for dealing with great volumes of data. Combining object-oriented and functional programming, Scala avoids bugs in complex applications with its static types, facilitates large-scale parallel processing, and, when paired with Apache Spark, provides high-performance cluster computing. Engineered to run on the JVM, Scala can run anything that Java runs. It’s becoming especially popular for people building complex algorithms or performing large-scale machine learning. Scala does feature a steeper learning curve than some other programming languages, but its massive user base is a testament to the value of sticking with it.

Julia

A much newer programming language than the others on this list, Julia has nevertheless made a strong impression thanks to its simplicity, readability, and lightning-fast performance. Designed for numerical analysis and computational science, Julia is especially useful for solving complex mathematical operations, which explains why it’s becoming a fixture in the financial industry. It’s also becoming widely known as a popular language for artificial intelligence, one reason many large banks now use Julia for risk analysis. However, because the language is relatively young, Julia lacks the variety of packages offered by R or Python—for now.

MATLAB

Widely used in statistical analysis, this proprietary numerical computing language is helpful for Data Scientists dealing with high-level mathematical needs, including Fourier transforms, signal processing, image processing, and matrix algebra. MATLAB has become widely used in industry and academia thanks to its intensive mathematical functionality. MATLAB can also help cut down on the time spent preprocessing data and help you find the best machine learning models, regardless of your level of expertise. It also features some great in-built plotting capabilities, making it a valuable data visualization tool.