Life, Death, and Shopping
summary
Step-by-step purchase prediction implemented in PySpark that turns terabytes of retail clickstream into labeled examples by generating moving model run dates (multiply dataset, avoid user/date overlap) and aggregating histories with time-based window functions (sum/avg/std over exclusive or inclusive ranges). Uses TF‑IDF weighting to score event types and offers arrays plus n‑grams/skip‑grams to capture event order; outcome is a binary label like purchase within the next week, suitable for survival/churn framing.
AI-written summary · about
chapters (9)
- 0:00Introduction to event prediction
- 1:36Overview of concepts
- 3:21Classical ML and survival analysis
- 5:55Data aggregation methods
- 10:27Window functions in practice
- 14:42Feature engineering and joining
- 18:17NLP methods for event sequencing
- 23:23Advanced TF-IDF and array features
- 30:45N-grams and concluding remarks
related talks
27mLife, Death, and Shopping
35mSpotting trends and tailoring recommendations: PySpark on Big Data in fashion
49mUnderstanding Customer Journeys with Deep Learning
26mLarge-Scale Recommendation System with Python and Spark
53mMachine Learning in E‑commerce
02h 00mData Science Driven Business Insights with Python and PySp
22mFrontiers of Machine Learning from the Lens of Retail
39mProductionising Data Science at Scale