Recommender System with Apache Spark Apache Kafka and Elasticsearch

Recommendations

Info

Elasticsearch Delimited Payload Filter Term Vectors Raw vector Custom analyzer Term vector with payloads 1.2 ⋯ −0.2 0.3 0|1.2 ⋯ 3|-0.2 4|0.3
Elasticsearch Custom scoring function Scoring • Native script (Java), compiled for speed • Scoring function computes dot product by: § For each document vector index (“term”), retrieve payload § score += payload * query(i) • Normalizes with query vector norm and document vector norm for cosine similarity
Page 1 and 2: Nick Pentreath Nov / 14 / 16 Buildi
Page 3 and 4: Agenda § Recommender systems & the
Page 5 and 6: Recommender Overview Systems
Page 7 and 8: The Machine Reality Learning Workfl
Page 9 and 10: Data Modeling for Recommender Syste
Page 11 and 12: User and Item System Requirements M
Page 13 and 14: Anatomy of a Data model User Event
Page 15 and 16: Why Kafka, Spark & Elasticsearch?
Page 17 and 18: Why Spark? DataFrames § Events & m
Page 19 and 20: Kafka for Recommender Systems
Page 21 and 22: Write to Event Store Event store Sp
Page 23 and 24: Item Metadata Analytics Spark Strea
Page 25 and 26: Structured Status Streaming § Stil
Page 27 and 28: Collaborative Matrix Factorization
Page 29 and 30: Collaborative Loading Data in Spark
Page 31 and 32: Deploying & Scoring Recommendation
Page 33: Recommendation Can we use the same
Page 37 and 38: Elasticsearch We get search engine
Page 39 and 40: Monitoring & Feedback
Page 41 and 42: System Events Logging Recommendatio
Page 43 and 44: Scaling Model Scoring
Page 45 and 46: Scoring Increasing number of shards
Page 47 and 48: Scoring Locality Sensitive Hashing
Page 49 and 50: Demo
Page 51 and 52: Future Work • Apache Solr version
Page 53: Thanks! https://github.com/MLnick/e

Recommender System with Apache Spark Apache Kafka and Elasticsearch

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?