Which of these free data science courses give a certificate
The 42 courses on this page fall into three groups: courses whose certificate is free, university courses on edX and MITx that sell a verified certificate on top of a free audit track, and open courseware that awards nothing.
Free certificates from Harvard's CS50, IBM and freeCodeCamp
CS50's Introduction to Databases with SQL and CS50's Introduction to Programming with R are free through Harvard's OpenCourseWare, and scoring at least 70% on every problem set and the final project earns a free CS50 Certificate; the edX edition of each sells a verified certificate instead. IBM's Cognitive Class courses come with a free certificate: Python for Data Science, SQL and Relational Databases 101, Data Science 101, Machine Learning with Python, Data Visualization with Python and Deep Learning Fundamentals.
freeCodeCamp publishes its certificates at a public URL anyone can check: Relational Databases is part of its current curriculum, while Data Analysis with Python and Machine Learning with Python are archived coursework, no longer updated but still available. The free data analytics certification courses page gathers these by credential.
Paid certificates on edX and MITx
HarvardX, BerkeleyX and StanfordOnline courses on edX are free to audit for a limited time, with graded work and the certificate paid. The HarvardX data science series charges $149 per course for Probability, Linear Regression, Wrangling and Building Machine Learning Models, and $219 for R Basics. Berkeley's Data Exploration is $130; Stanford's Statistical Learning with Python is $186.
The MITx courses in the Statistics and Data Science MicroMasters, such as Fundamentals of Statistics and Machine Learning with Python: from Linear Models to Deep Learning, audit free and price the certificate track at $300, and are instructor-paced rather than self-paced.
Free with no certificate: OpenCourseWare, Khan Academy and Microsoft Learn
MIT OpenCourseWare and the Open Learning Library publish complete courses with no credential, including Introduction to Machine Learning, Gilbert Strang's Matrix Methods in Data Analysis, Signal Processing, and Machine Learning and Statistics for Applications. Stanford Engineering Everywhere's CS229 Machine Learning is free and unregistered. Khan Academy's Statistics and probability has no paid tier and no certificate. Microsoft Learn paths are free to read and award no certificate on their own.
What Harvard's free data science courses cover: the R series and Python
Harvard's R series is meant to be taken in order: R Basics introduces the language through a single dataset on crime in the United States at 1 to 2 hours a week over eight weeks; Probability teaches random variables, Monte Carlo simulation and the Central Limit Theorem through the financial crisis; Linear Regression uses the Moneyball case to explain regression and confounding; Wrangling covers importing, scraping, regular expressions and dplyr; and Building Machine Learning Models has you build a movie recommendation system.
For Python, Introduction to Data Science with Python covers regression and classification with pandas, NumPy and scikit-learn in eight weeks, and Machine Learning and AI with Python goes from decision trees to random forests; both are rated intermediate and assume some Python. Introduction to Probability and Statistics and R supply the mathematical side.
Which free machine learning courses are here, from concept videos to MIT's class
IBM's Deep Learning Fundamentals is concept-only video, about 3 hours with no prerequisites, and its Machine Learning with Python surveys K-nearest neighbours, decision trees, regression and clustering in Jupyter notebooks over about 20 hours, with Python for Data Science as its prerequisite.
At the rigorous end, MIT's Introduction to Machine Learning recommends Python, calculus and linear algebra and asks about 12 hours a week for thirteen weeks; Machine Learning with Python: from Linear Models to Deep Learning runs at the pace of the campus class over fifteen weeks; and Introduction to Deep Learning needs calculus at the level of derivatives and linear algebra at the level of matrix multiplication. Andrew Ng's CS229 asks for programming, basic probability and basic linear algebra. Only the IBM and freeCodeCamp machine learning courses give a free certificate; the MIT and Stanford courses either sell one or issue none.
Where to start learning data science free: SQL, Python or statistics
If you have never written code, IBM's Python for Data Science is the entry point, about 18 hours from first program to Pandas and NumPy in a browser-based lab. For databases, SQL and Relational Databases 101 takes about 5 hours and lists no prior skills, while CS50's Introduction to Databases with SQL goes deeper over seven weeks at 6 to 12 hours a week, from SQLite to PostgreSQL and MySQL, with no prior programming assumed.
For statistics without code, Khan Academy's Statistics and probability runs from displaying data to inference in 16 units with practice and unit tests. To test whether the field is for you, IBM's Data Science 101 is an orientation course built on practitioner interviews with no coding, and Microsoft's Introduction to Microsoft Azure Data core data concepts explains data roles and workloads in about 1 hour.
How much mathematics and programming the courses assume
Most IBM courses, the Harvard R series and the Microsoft paths are rated beginner. Berkeley's Fundamentals of Economics is classed introductory by edX but lists Data 8X as a prerequisite. Stanford's Mining Massive Datasets is the only listing rated advanced, intended for graduate students with algorithms, databases, linear algebra, calculus and statistics behind them. The MITx MicroMasters courses expect calculus, and Fundamentals of Statistics is explicitly not a first course in data analysis. The free online mathematics courses and free online python courses pages cover those foundations.