Explore best practices for building a recommender system using Apache Mahout and Solr in this 57-minute conference talk from PhillyETE2014. Discover effective data selection techniques, learn how to leverage Apache Mahout for co-occurrence discovery, and implement an innovative search technology approach. Gain insights into deploying recommenders with Apache Solr for rapid response in production environments. Examine the machine learning workflow, recommendation engines, and data processing techniques. Delve into topics such as history matrices, co-occurrence matrices, and indicator matrices. Learn about setting up Solr indexes, offline analysis, and real-time recommendations using MapR data platform. Explore the advantages of dithering and architectural considerations. Get an overview of Apache Mahout, its integration with Scala, Spark, and h2o, and learn about the project's roadmap and how to get involved in this open-source initiative.
Building a Recommender: Best Practices with Apache Mahout and Solr - Emerging Technologies 2014
Overview
Syllabus
EMERGING TECHNOLOGIES FOR THE ENTERPRISE CONFERENCE
What Does ML Look Like?
Learn a New Language
What Do You See?
Human insight plays an important role
Typical Machine Learning Workflow
Recommendation: Widely Used Machine Learning
Get Useful Indicators from Behaviors
History Matrix: Users by Items
Co-occurrence Matrix: Items by Items
Indicator Matrix: Anomalous Co-Occurrence
Setting up Solr index: Item Meta-Data
Welcome to the Music Machine!!
Sample Music Log File Data
Offline Analysis
Internals of the Recommender Engine
Real-time recontmendations using MapR data platform
A Quick Simplification
Architectural Advantage
Better Long-Term Recommendations
Why Use Dithering?
Apache Mahout: Overview
Mahout and Scala
Mahout and Spark
Mahout and h2o
Roadmap: Apache Mahout 1.0
Join Apache Mahout
Learn about Apache Mahout
Taught by
ChariotSolutions
