Python is by far the most commonly used programming language for IBM data science capstone projects on GitHub. Python has become the dominant language for data science due to its rich ecosystem of packages and libraries for data wrangling, analysis, visualization, and machine learning. Key Python data science libraries like Pandas, NumPy, Matplotlib, Seaborn, scikit-learn, Keras, and Tensorflow are ubiquitously used across projects. Python’s clear and readable syntax also makes it very approachable for newcomers to data science. Many capstone projects involve analyzing datasets from a variety of domains using Python for tasks like data preprocessing, exploratory data analysis, building predictive models, and creating dashboards and reports to communicate findings.

R is another popular option, especially for more statistics-focused projects. R’s strengths lie in implementing statistical techniques and modeling, and it includes powerful packages like ggplot2, dplyr, and caret that are very useful for data scientists. While Python has gained more wide adoption overall, R still maintains an active user base in fields like healthcare, finance, marketing that involve intensive statistical analysis. Some IBM data science capstones apply R for predictive modeling on tabular datasets or for time series forecasting problems. Data visualization is another common application thanks to R’s graphics capabilities.

JavaScript has increased in usage over the years and is now a viable language choice for front-end data visualization projects. D3.js in particular enables creation of complex, interactive data visualizations and dashboards that can be embedded within web pages or apps. Some capstones take JSON or CSV datasets and implement D3.js to build beautiful, functional visualization products that tell insightful stories through the data. JavaScript’s versatility also allows integration with other languages – projects may preprocess data in Python/R and then render results with D3.js.

SQL (often SQLite) serves an important role for projects involving relational databases. Even if the final analysis is done in Python/R, an initial step usually involves extracting/transforming relevant data from database tables with SQL queries. Healthcare datasets in particular are commonly extracted from SQL databases. SQL knowledge is invaluable for any data scientist working with structured datasets.

Most machine learning engineering capstones will involve some use of frameworks like TensorFlow or PyTorch when building complex deep learning models. These frameworks enable quick experimentation with neural networks on large datasets. Models are trained in Python notebooks but end up deployed using the core TensorFlow/PyTorch libraries. Computer vision and NLP problems especially lend themselves to deep learning techniques.

Java is still prevalent for projects requiring more traditional software engineering skills rather than pure data science. For example, building full-stack web services with backend APIs and database integration. frameworks like Spark and Hadoop see usage as well for working with massive datasets beyond a single machine’s resources. Scala also comes up occasionally for projects leveraging Spark’s capabilities.

While the above languages dominate, a few other options do come up from time to time depending on the specific problem and use case. Languages like C/C++, Go, Swift may be used for performance-critical applications or when interfacing with low-level system functionality. MATLAB finds application in signal processing projects. PHP, Node.js, etc. can be applied for full-stack web/app development. Rust and Haskell provide quality alternatives for systems programming related tasks.

Python serves as the most popular Swiss army knife for general data science work. R maintains a strong following as well, especially in domains requiring advanced statistical modeling. SQL is ubiquitous for working with relational data. JavaScript enables data visualization. Deep learning projects tend to use TensorFlow/PyTorch. Java powers more traditional software projects. The choice often depends on the dataset, goals of analysis, and any specialized technical requirements – but these programming languages cover the vast majority of IBM data science capstone work on GitHub. Mastering one or two from this toolkit ensures data scientists have the tools needed to tackle a wide range of problems.