poi_recommendation.md

POI推荐

场景: yelp, 美团, airbnb

• Design a system to find nearby restaurants
• Design a system to match drivers with riders for Uber
• Design a system to compute ETA for food delivery

特点: 如果是event 推荐这种注重实效性、位置性的推荐，event发生后不存在了，所有item可以认为都是冷启动 对于位置的挖掘可采用图特征或模型

1. requirements

products/use cases

objective

• connect people with great local businesses

constraint

• data
• volume
• latency

2. ML task & pipeline

预测目标

• 是否点击
• 停留时间(dwell time), 可转化为t/(t+1)来逼近sigmoid函数，t很大时接近1；很小时接近0

3. data collection

• user
  • User location: For localized recommendations we need to consider only businesses near the city or neighborhood where the user is located

4. feature

• sparse
• dense

5. model

retrieval

• 取决于filter

ranking

rerank

6. evaluation

• offline
  • NDCG
  • MAP
• online: A/B testing holdout canary

7. deploy & serving

• batch serving
• online serving

8. monitor & maintenance

9. 优化与问答

冷启动的item

• 双塔可以采用default embedding, 而不是random initial

reference

• yelp-Beyond Matrix Factorization: Using hybrid features for user-business recommendations
• Yelp Food Recommendation System
• 美团-旅游推荐系统的演进
• 美团-基于机器学习方法的POI品类推荐算法
• Design Yelp
• System Design — Nearby Places Recommender System