One workshop that data scientists don't want you to attend...
summary
Shows how a real system is built and what actually works: labels about 1,500 labeled articles, vectorizes titles+descriptions with pandas and a scikit-learn pipeline, trains an SVM using count/TF‑IDF features and grid search to reach about 96.5% accuracy on held‑out data, and exports a pickled model behind a simple NewsAPI/Flask web service that classifies live headlines (BuzzFeed scores highly for clickbait).
AI-written summary · about
chapters (9)
- 0:00Introduction
- 12:19Scikit-learn transformers & classifiers
- 22:45SVM for clickbait classification
- 32:37Train-test split & evaluation
- 45:24Feature engineering for performance
- 53:20Hyperparameter tuning with grid search
- 1:00:39Model versioning with pickle
- 1:11:19Web service for clickbait detection
- 1:17:07Live demo & conclusion
related talks
34mApplying Machine Learning Techniques to Detect and Understand Fake News
02h 19mWorkshop: Demystifying Natural Language Processing using Python (Scikit-Learn/ Keras)
04h 02mHands-on Session on Data Science for Beginners
40mEveryone can Data! - A Practical Data Science Intro
44mTeaching Computers ABCs: A Quick Intro to Natural Language Processing- 34mApplying Machine Learning Techniques to Detect and Understand Fake News
48mHow to make intelligent web-apps
38mData science at The New York Times