Professional Documents
Culture Documents
Contextual Markov Models
Contextual Markov Models
Contextual Markov Models
7 December 2013
DDDM@ICDM2013,
Dallas, TX, USA
Outline
• What is predictive Web analytics
• Context-Aware Predictive Analytics framework
• User intent modeling
• Contextual Markov Models
• Case study, experimental results
• Conclusions and further ongoing work
y labels 2. application
label? y'
DDDM@ICDM2013 Predicting Current User Intent with Contextual Markov Models 6
Dec 7, 2013 Mykola Pechenizkiy, Eindhoven University of Technology
Context in IR & RecSys
• User Context
– Preferences, usage history, profiles
• Document/Product Context
– Meta-data, content features
• Task Context
– Current activity, location etc.
• Social Context
– Leveraging the social graph
Time of
the day
context
no context
Male
don’t buy
gender Female
context don’t buy
relevance
relevance
DDDM@ICDM2013 Predicting Current User Intent with Contextual Markov Models 9
Dec 7, 2013 Mykola Pechenizkiy, Eindhoven University of Technology
Context-Awareness as Meta-learning
Training:
population
Test
data
??
X'
labels
Environment/
Context
Application:
Historical
y' = Lj (X')
X Model L
data Lj <= G(X',E)
y labels
label? y'
DDDM@ICDM2013 Predicting Current User Intent with Contextual Markov Models 10
Dec 7, 2013 Mykola Pechenizkiy, Eindhoven University of Technology
Learning Classifiers & Context
Quick
Search
Universities
in the
spotlight
Banner
Click
Mastersportal.eu - Search
Refine
Search
Click on Program
is Search Result
Click on Click on
University Country
User Navigation Graph
Motivation for Contextual Markov Models
Useful Contexts:
E[M] < pc1*E[Mc1] + pc2*E[Mc2]
Why should it help?
Discover clusters in
the graph using
community
detection
algorithm
user location
user type
Refine Product
Search Click Payment Search Click
Search View
t0 t1 t2 t3 t4 t5 t6
Timeline
What is next?
Change of intent?
Collaboration is welcome!
DDDM@ICDM2013 Predicting Current User Intent with Contextual Markov Models 25
Dec 7, 2013 Mykola Pechenizkiy, Eindhoven University of Technology
Designing Context-awareness
Context-aware Adaptation
Feature set expansion
Model selection/weighting
Feature set selection
Model adjustment Output correction
Instance set selection
Predictive
Training data Predictions
model(s)