Ranking Listings with Machine Learning and ElasticSearch
summary
Describes how cian.ru implemented personalized ad ranking for search by combining ElasticSearch with complex machine‑learning models (gradient boosting and neural networks). Explains architecture and engineering approaches used to apply these models efficiently at scale to hundreds of thousands of listings, overcoming ElasticSearch’s lack of built‑in personalized ranking.
AI-written summary · about
chapters (8)
- 0:00Company overview
- 5:07Current ranking problem
- 10:26Ranking task definition
- 16:04User-item matrix
- 21:30Data pipeline
- 27:13Model training & evaluation
- 31:09Business impact & future
- 37:02Q&A
related talks
44mRobust AI Search Ranking for Radical C2C Marketplace Growth
47mLearning to rank search results
55mLearning to Rank, a Little
38mLearning to Rank: where search meets machine learning #bbuzz
51mHow to Effectively Rank All of Runet's Products
41mDynamic Ranking of Search and Recommendation Results in High-Traffic Systems
25mBetter search relevance using Learning to Rank at mobile.de
21mAmazon Search: The Joy of Ranking Products