In the report we share the experience of creating a content-based Recommender system for electronic Commerce on the semantic core of the Runet. Describe how organized centralized collection and processing of information about the users visit more than 100,000 sites of various kinds on the basis of Amazon Kinesis. Share the experience of multithreaded online indexing of data streams in Lucene. Demonstrate the underlying ranking algorithms and the formation of personal recommendations to visitors over 20 000 online shops.

Aleksandr Serbul

The Head of control and quality of integration and implementations, 1C-Bitrix