Source-linked AI summary
Curriculum Guidelines for Undergraduate Programs in Data Science
Richard De Veaux, Mahesh Agarwal, Maia Averett, Benjamin Baumer, Andrew Bray, Thomas Bressoud, Lance Bryant, Lei Cheng, Amanda Francis, Robert Gould, Albert Y. Kim, Matt Kretchmar, Qin Lu, Ann Moskol, Deborah Nolan, Roberto Pelayo, Sean Raleigh, Ricky J. Sethi, Mutiara Sondjaja, Neelesh Tiruviluamala, Paul Uhlig, Talitha Washington, Curtis Wesley, David White, Ping Ye
TL;DR
The paper addresses how institutions can structure undergraduate Data Science majors amid rapid program growth and continuing debate about the field. It develops curriculum guidelines centered on integrated, data-driven learning across mathematics, computer science, and statistics, with communication and data management included. The resulting framework is a flexible blueprint whose authors expect to revise it as the field evolves.
Problem
Rapid growth and continuing debate about Data Science leave institutions needing structure for undergraduate programs and major revision.
Method
The authors synthesize faculty perspectives into guidelines for an integrated curriculum spanning mathematics, computer science, and statistics, with data experiences across the curriculum.
Results
The paper proposes a curriculum blueprint combining foundational knowledge, the recursive data cycle, communication, data curation, and interdisciplinary coursework.
Takeaways & Limitations
Institutions can use the guidelines to plan or revise a Data Science major while adapting the proposed integrated model to local course constraints.
Takeaways & Limitations
The guidelines are a first attempt for an evolving field and will require periodic revision.
Abstract
from arXiv · showhide
The Park City Math Institute (PCMI) 2016 Summer Undergraduate Faculty Program met for the purpose of composing guidelines for undergraduate programs in Data Science. The group consisted of 25 undergraduate faculty from a variety of institutions in the U.S., primarily from the disciplines of mathematics, statistics and computer science. These guidelines are meant to provide some structure for institutions planning for or revising a major in Data Science.
1. Introduction
The guidelines respond to rapidly expanding Data Science education by proposing an interdisciplinary undergraduate curriculum grounded in mathematics, statistics, and computer science. They emphasize integrated, data-centered learning and course redesign rather than simply assembling existing offerings.
- Motivation: Data Science programs expanded rapidly as complex data proliferated and anticipated workforce demand grew.The paper notes 530 programs in related fields at more than 200 universities, mostly at the master's and certificate levels.
- Guideline Development: The PCMI workshop brought together 25 faculty from mathematics, statistics, and computer science to develop undergraduate curriculum guidelines.The participants represented liberal arts colleges and research universities.
- Scope: The guidelines are intended as an informative blueprint rather than a prescriptive standard, and the authors expect revision as Data Science evolves.They also state that the proposed mathematics pathway is not an alternative route for mathematics majors without additional theoretical foundations.
- Course Design: Integrated course redesign could reduce the major from 14 or more courses using existing offerings to fit a typical 10-course liberal arts major.The proposed redesign aims to improve efficiency and exploit synergies among mathematical, computational, and statistical content.
- Interdisciplinary Curriculum: The proposed major is interdisciplinary, combining mathematical foundations with computational and statistical thinking rather than treating the disciplines as separate.The paper describes this integration as central to Data Science problem solving.
- Data-Centered Learning: Students should practice the full recursive data cycle, from obtaining and managing data through analysis and communicating results, using real-world projects.The guidelines place data experiences throughout the curriculum and emphasize raw, varied data and project-based applications.
3. Key Competencies and Features of a Data Science Major
A Data Science major integrates computational, statistical, and mathematical thinking to prepare students to work with data throughout an investigation. Graduates should manage data, build and assess models, solve problems algorithmically, and apply these skills in contextual and team-based settings.
- Data Science graduates should work with data at all stages of an investigation and within team environments.
- Statistical preparation covers data analysis, collection, modeling, inference, and the limits imposed by data issues and model assumptions.
- Computational preparation includes databases, web scraping, text processing, secure storage, programming, algorithms, and software-performance considerations.
- The major synthesizes computational and statistical thinking practices while relying on mathematical foundations and connections among these domains.
- Students should choose, fit, use, build, and assess mathematical, statistical, and machine-learning models, using visualization to communicate results and identify weaknesses.
- Programs should include investigations in complementary disciplines because data arise from particular contexts that are necessary for analysis.
1. Communication
Communication, ethical practice, and integrated technical coursework are central features of the proposed curriculum. The guidelines favor repeated, data-centered experiences that connect programming, mathematics, statistics, databases, and applied projects.
- 1. Communication: Effective communication is a core data-science skill across teammates, less-specialized audiences, and the public.Students should practice oral, written, and visual communication throughout the curriculum.
- 1. Communication: Data Science programs should integrate ethical training on data ownership, security, privacy, transparency, and reproducibility.
- The curriculum repeatedly engages students in the full cycle of learning from data through interwoven and integrated topics and tools.
- 4.1. Overview of Course Sequence: The proposed sequence includes introductory Data Science, mathematics, algorithms and software, databases, statistical modeling, machine learning, an outside-discipline course, and a capstone.
- 4.1. Overview of Course Sequence: Introductory courses use a high-level language to explore, visualize, and question data before introducing a more algorithmic language.
- 4.1. Overview of Course Sequence: Algorithmic foundations cover problem decomposition, efficient solutions, programming techniques, data structures, and the performance of software and packages.
- 4.1. Overview of Course Sequence: Data management must address relational and NoSQL systems, web services, Internet-accessible data, versioning, provenance, and scaling to distributed or streaming data.
5. Additional Considerations
The guidelines identify boundaries around graduate preparation and community-college articulation. Existing institutional resources may require collaboration and additional preparation before transfer into a Data Science degree.
- 1. Graduate Study: The proposed mathematics courses are unlikely to prepare students for immediate acceptance into mathematics, statistics, or computer-science PhD programs.
- 2. Articulation with community colleges: Community-college students can prepare for transfer with calculus, introductory computer science, and data-oriented introductory statistics.
- 2. Articulation with community colleges: Some community colleges may lack resources for the proposed Data Science I and II courses, making cross-department collaboration important.
- 2. Articulation with community colleges: Statway and New Mathways sequences may provide statistical and basic computational thinking depending on local implementation.
3. Prerequisites and preparation in high school
Preparation for Data Science includes statistical, mathematical, and computational foundations before or alongside college study. The guidelines emphasize calculus readiness, matrix algebra, and basic data-analysis concepts, with applied experiences continuing through the curriculum.
- 3. Prerequisites and preparation in high school: Incoming students should be calculus ready, including a strong precalculus background.
- 3. Prerequisites and preparation in high school: Basic matrix algebra, including solving systems of linear equations, is helpful preparation.
- 3. Prerequisites and preparation in high school: Students should understand scatterplots, correlation, lines of best fit, and descriptive statistics such as center, spread, and skew.
- 4. Internship and applied experiences: Internships and applied experiences are significant parts of the program, with practical projects central to the capstone course.
6. Transitioning to a Data Science Major Using Typical Existing Courses
The proposed major uses an integrated ten-course foundation spanning mathematics, computer science, and statistics, while recognizing that existing-course pathways may require more coursework and eventual redesign.
- The curriculum’s ten highlighted courses provide the bare minimum basic skills for a Data Science major.Programs may add courses or adopt ten newly designed integrated courses.
- The proposed foundation integrates topics from mathematics, computer science, and statistics.The authors present these disciplines as the basis for the major at present.
- Mathematics courses should connect mathematical methods to real-world, especially data-driven, problems.
- The guidelines are intended as a starting point for building new programs and transitioning existing ones, not as prescriptive requirements.
- The authors identify faculty development, coordination with two-year colleges and high schools, and periodic revision as issues requiring continued attention.The field’s rapid evolution makes revisions necessary.
8. Appendix – Detailed Courses for a Proposed Data Science Major
The appendix proposes courses organized around the full data workflow, combining programming, data curation, analysis, modeling, communication, teamwork, and ethics with increasingly comprehensive projects.
- Introduction to Data Science: The proposed introductory course covers the full data workflow from data acquisition to communicating final results.It is described as a high-level introduction taught through an iterative cycle and collaborative projects.
- Introduction to Data Science: Students practice exploring and wrangling data, writing basic functions, summarizing and visualizing data, and modeling deterministic and stochastic phenomena.
- Introduction to Data Science: The curriculum includes written, oral, and graphical communication, ethics, teamwork, version control, and application-motivated projects.
- Data and Data Curation: Students encounter varied data types and sources, including relational databases, web/API data, and streaming data, while learning data curation and structure-building.
- Data and Data Curation: Data collection, cleaning, extraction, exploratory analysis, estimation, inference, and model assessment are treated as connected course components.Topics include sampling design, wrangling, SQL, confidence intervals, simulation, resampling, and training/testing sets.
- Data and Data Curation: A comprehensive data-driven project requires problem formulation, informed wrangling, elementary analysis, conclusions, and attention to generalizability limitations.The outlined experiences are also associated with enhanced ability to collect data and better tools for analysis and scope of inference.
3. Mathematics for Data Science I (a) Vision
Mathematics for Data Science I emphasizes application-driven linear modeling and optimization, using computation and geometric intuition rather than proof-centered or hand-calculation-heavy instruction.
- Vision: The course introduces linear modeling and optimization through mathematical foundations motivated by applications.Numerical computation connects the concepts to real-world ideas.
- Modeling and Optimization: Students model real-world phenomena with functions and learn to optimize functions and interpret the results.
- Linear Algebra and Geometry: The curriculum develops geometric intuition for linear modeling, optimization, correlation, and higher-dimensional models.Topics include vectors, projections, matrix algebra, systems of equations, and geometric interpretations.
- Modeling and Optimization: Students investigate how models respond to changes in inputs and use tools for understanding local model behavior.
- Modeling and Optimization: Students must navigate model choice while understanding model limitations.
- Linear Algebra and Geometry: The course includes solving linear systems and optimization in the context of Data Science and regression.
4. Mathematics for Data Science II (a) Vision
Mathematics for Data Science II introduces probability through data-generating processes, empirical distributions, visualization, and numerical computation, alongside computational problem solving and algorithmic efficiency.
- Vision: The course builds probability distributions from empirical data and density functions from histograms, emphasizing visualization and numerical computation.
- Probability: Students learn to recognize randomness in data-generating processes and apply probability properties to specific problems.
- Probability: Students use simulation to generate data and explore probability, including sampling distributions and the central limit theorem.
- Probability: The curriculum covers simulation, conditional probability, independence, common distributions, sampling distributions, and multivariate extensions.
- Computational Problem Solving: Computational problem solving deepens through data-centered procedural programming, problem-solving strategies, and efficiency assessment.
- Computational Problem Solving: Students write complete programs, learn a second programming language, and understand implementations and efficiency of basic statistical algorithms.
6. Data Curation, Management, Organization (a) Vision
The guidelines emphasize acquiring, transforming, organizing, and assessing data from diverse sources, including large-scale and unstructured systems, so it can support exploration, visualization, and analysis.
- Vision: Data curation should span single and multiple sources, varied structures, real-world data, transformation, wrangling, and big-data management.The curriculum also includes advanced cleaning, tool integration, and query operations.
- Vision: Students should acquire data ranging from schema-based systems to unstructured sources requiring APIs, scraping, munging, or distributed storage.They should also organize and unify data from multiple sources and generalize their skills to new APIs and systems.
- Vision: Students should understand data-source bias and randomness, including how these issues affect analysis and generalization.The guidelines place these concerns alongside authentication, authorization, and ethical access to remote data.
- Vision: Students should understand the data structures needed for exploration, visualization, and analysis and transform unsuitable data into appropriate forms.This objective connects acquisition and management skills directly to downstream analytical use.
7. Intro to Statistical Models
The introductory statistical-models course builds inference around regression, beginning with linear models and extending to nonlinear alternatives, uncertainty, and model selection using real data.
- Vision: The course introduces a regression-based framework for inference, using linear models as the foundation and comparing them with nonlinear approaches.The material is grounded in real-world data and connects common structures across statistical and machine-learning methods.
- Learning goals: Students should understand the mathematical framework for linear models and apply regression to real datasets for inference, understanding, and communication.The intended application is scientific inference and communication rather than model fitting alone.
- Learning goals: The course includes theoretical and simulation approaches to uncertainty, hypothesis testing, confidence intervals, estimation, and likelihood or Bayesian principles.These topics support interpretation of regression results and uncertainty in models.
- Learning goals: Model-selection topics include regularization, the bias/variance tradeoff, parsimony, AIC, BIC, ridge regression, and LASSO.These methods provide alternatives for controlling model complexity and selecting among candidate models.
8. Statistical and Machine Learning (a) Vision
The guidelines frame statistical and machine learning as complementary perspectives, combining theoretical understanding with practical application to real problems and culminating in an integrated capstone experience.
- Vision: The course blends computer science’s algorithmic perspective with statistical thinking’s predictive perspective and applies common methods across disciplines.Students are expected to understand theoretical foundations and develop practical skills for new problems in science and industry.
- Learning goals: Students should learn widely used algorithms, understand their theoretical underpinnings and inference limitations, and evaluate and communicate performance on real data.The guidelines also emphasize understanding algorithm inputs, outputs, and solvable problem types.
- Topics: The curriculum covers transformations, dimension reduction, smoothing, regression, classification, clustering, ensemble methods, model selection, assessment, and performance measures.Examples include PCA, regression trees, logistic regression, k-means, boosting, bagging, and cross-validation.
- Capstone: The capstone integrates knowledge from mathematics, statistics, and computer science through a real-world project spanning collection, processing, analysis, solution development, and communication.Projects may be application-focused, algorithmic, computational, or theoretical, with students working in teams and presenting reproducible, ethical results.