// the find
justmarkham/DAT4
General Assembly's Data Science course in Washington, DC
Course materials from a General Assembly data science bootcamp that ran December 2014 through March 2015. Covers the classical ML curriculum — Python, pandas, scikit-learn, KNN, logistic regression, decision trees, ensembling, NLP, clustering, recommenders — in 22 class sessions. Aimed at beginners learning applied ML with Python.
The README doubles as a complete syllabus with timestamped class notes, homework assignments, and resource links — unusually thorough for a course repo. Code examples are standalone scripts rather than monolithic notebooks, which makes individual topics easy to extract. The bias-variance and model evaluation sections include concrete exercises with solution files, not just lecture slides. Coverage is genuinely broad for 10 weeks: SQL, MapReduce, NLP, and recommenders all get dedicated sessions with working code.
Frozen in 2014: Python 2.7, pre-1.0 scikit-learn APIs, links to Codecademy and Coursera courses that no longer exist or have moved. The Pandora scraping code is almost certainly broken. No tests, no reproducible environment (no requirements.txt, no conda env file), so you cannot actually run most of this without chasing down compatible versions. The NLP session leans heavily on NLTK and textblob, which have both aged poorly relative to spaCy or Hugging Face — a student following this in 2025 would be learning the wrong tools.