01/ 01
- Event collection: impressions, clicks, purchases, dwell time, likes, shares - Feature store: online (Redis / Feature Store like Feast) and offline (parquet on S3) - User features: demographics, history, preferences, recency-weighted interests - Item features: category, tags, embeddings, popularity, freshness - Interaction features: cross of user-item attributes - Session-based features: real-time click stream aggregation (Flink / Spark Streaming)
自动翻译
自动翻译中…