Vaclav Petricek - Hadoop Summit 2013 - theCUBE - #HadoopSummit
summary
eHarmony describes a concrete pipeline: three modeled problems — long‑term compatibility, affinity (mutual interest), and distribution (who and when) — trained on 10+ years of historical interactions and thousands of married couples. Data includes a 150‑question questionnaire, behavioral signals (logins, messages, device), and geographic/age filters. Data lives in HDFS + Hive; models use Vowpal Wabbit at scale and sometimes genetic algorithms for feature search, and optimize for mutual responses and increased marriages.
AI-written summary · about
