Using User Behavior for
Goal: Revenue Optimization with User Behavior Data
● My blog: http://blog.trieu.xyz
● Creator and Founder at RFX Lab http://www.rfxlab.com
1. Introduction to the User Behavior dataset
2. Why does User Behavior impact revenue?
3. How to build User Behavior database?
4. How to target ads with User Behavior database?
5. Research Plan & Timeline Milestone
The goal of a behavioural targeting system is to
automatically decide which creatives are most relevant to
be presented to a web user visiting a web page based on
the previous recorded on-line behaviour of that user.
E.g: a user who is exceptionally frequently visiting sport-
related web pages could be presented more sport-
oriented ads than others.
Why does User Behavior impact revenue?
Revenue ~ Sum(click CPC) + Sum(true-imp CPM)
Revenue Optimization for CPC
=> improve the probability of user click
=> improve the relevance of banner and user interest in
Revenue Optimization for CPM
=> avoid wasting impression (out of banner)
An user could click this banner if
● a male
● a Samsung fan (e.g: like Samsung fan page )
● a geek fan (read a lot of tech reviews)
● age: 16 - 30
● need a new smartphone
● more than 3 page visits of Fshop / Sendo / Lazada
● live in Vietnam
● Tracking API for client (Web + Mobile apps)
● Collecting user behavior logs from Kafka
● Design & implement User Behavior Profile with HBase
● Design query API for Ad Delivery
● Real-time Revenue Optimization System with Spark
● Design core master algorithms
● Tracking Users on the Internet with Behavioral Patterns
● Learning to Target: What Works for Behavioral Targeting
● Behavioural Targeting in On-Line Advertising
● Anti Click Spam
● Real-time User Segmentation using HBase
● Behind the Scenes of Behavioral Advertising
● Extracting User Behavior