Lucene/Solr Revolution 2017 has ended
View analytic
Friday, September 15 • 10:30am - 11:10am
R to forecast Solr activity
This session is a deep review on Solr performance management and how to set up a scalable Solr infrastructure that will match with future user's activity.

Taking advantage of solr logs and time series functions in R, we can build custom models to analyze Solr activity and highlight periodicity. Using other kinds of R functions, such as exception management, we can highlight activity peak and keep it or not in the predictive model. Using the Association function in R, we can analyze user search behavior (such as cascading search, one search criteria leading to another facet of search).

Once a user's activity is statistically validated (exploration and discovery techniques using Dashboard, Olap, etc.), we can develop custom predictive models in R to forecast user activity and adapt Solr infrastructure to this expected workload.

The last piece of the framework is to use a comparison model between forecast and reality to adjust the custom model and address machine learning points of interest.

avatar for Patrick Beaucamp

Patrick Beaucamp

CEO, Bpm-Conseil
Patrick Beaucamp is founder of the Vanilla project, the only true Open Source Business Intelligence Platform, and Chairman of Bpm-Conseil, the company behind the Vanilla project. | | Patrick is a regular speaker at the Open Source Conference to talk about data visualisation, business... Read More →

Friday September 15, 2017 10:30am - 11:10am
South Seas B