Is Bigger Really Better? Predictive Analytics with Fine-grained Behavior Data
summary
Case study compares models built from 200–300 sociodemographic and geographic variables versus models that add one million customers described by 3 million merchants of fine‑grained transaction behavior; predictive lift plateaus quickly with traditional features but continues rising — increasing returns to scale — as more behavioral examples are added, seen for a savings product and a pension product, implying sparse, high‑cardinality behavior streams materially boost predictive performance.
AI-written summary · about
chapters (4)
- 0:00Start
- 1:55Is Bigger Really Better? Predictive Analytics with Fine-grained Behavior Data
- 2:48Learning curve shows no improvement in lift with massive data for targeting a banking product
- 6:43Using fine-grained behavior data, learning curves show continued improvement to massive scale
related talks
56mPredictive Analytics for Practitioners
40mMachine Learning over Small Data: generating new variables with predictive power
27mUnlocking Product Growth With Big Data, Data Science And Machine Learning
15mUnderstanding Decisions Driven by Big Data
38mDo's and don'ts of big data analytics
24mUnlocking revenue: newcomer classification for boosted profits
58mNew-type Scoring Models: Analyzing Customer Actions
01h 39mApplying Rigorous Machine Learning Methods in Business Strategy