Generative AI for Social Good at Open Data Science East 2024
Context Mining and Integration in Web Predictive Analytics
1. Context mining and integration into
Predictive Web Analytics
Julia Kiseleva (Eindhoven University of Technology),
Supervised by:
Mykola Pechenizkiy (Eindhoven University of Technology),
2. Web Predictive Analytics
What is predictive web analytics?
Web predictive analytics:
• aims to predict individual and aggregated characteristics
indicating visitor behavior for purposes of understanding and
optimizing web usage.
• Application:
o Search engines
o Recommender System
• Examples:
o Computational Advertisement
• Predictive web analytics tasks:
o Online shop’s recommendations;
o Users’ next action prediction;
o Users’ intention predicting;
o Personalized search result page.
3. Model L
Users web log
Historical
data
labels
label?
1. training
3. application
X
y
X'
y’=L (X')
Formulations:
① Classification
② Regression
③ Clustering
④ Scoring
labels
Testing
data
2. testing
Predictive Web Analytics
4. User next action prediction
Historical data. Actions ={Search, Refine Search, Click
on Banner, Product view, Payment}
Search
Refine
Search ?Click on
Banner
Product
View
What is next?
Session 1 Search Refine Search Click on Banner Product View Payment
Session 3 Product View Payment
Session 3 Search Refine Search Refine Search Click on Banner
Session 4 Search Refine Search Click on Banner Product View Payment
Session 5 Product View Click on Banner Search
Running Example: users’ trail predictions
5. Search
Refine
Search ?Click on
Banner
Product
View
What is next?
Running Example: users’ trail predictions
Search
Refine
Search
Payment
Click on
Banner
Product
View
1.0 2/3
1/3
1/2
1/4
Drop out
3/4
1/4
1
1/4
User next action prediction
6. Context
What is context? – any additional information that
Why we need context?
o enhances the understanding of the instance of interest,
o helps us to classify this instance or makes predictions regarding its
behavior.
• Two major context types:
o Explicit – stored explicitly or given by domain expert (location, OS,
Browser)
o Implicit – hidden in the data. We need techniques to discover context.
7. Taxonomy for explicit
Context
Human Factors
Physical
Environment
Factors
User
Characteristics
Social
Environment
Intent
Conditions
Infrastructure
Location
*Weather
*Light
*Acceleration
*Audio
*…
*Temperature
*Humidity
*…
8. Environment/
Context
Model L
Users web log
X'
y'
Historical
data
labels
X
y
label?
labels
Test
data
Strategies:
① ?
② ?
③ ?
④ ?
Context-Awareness in
Web Predictive Analytics
9. Research Questions
Question 1: How to define the context in predictive web
analytics?
Question 2: How to connect context with the prediction
process in predictive web analytics?
Context
Definition
Context
Discovery
Context
Modeling
Context Mining:
How define context?
Context
Integration
10. Search
Refine
Search ?Click on
Banner
Product
View
What is next?
Running Example: users’ trail predictions
Search
Refine
Search
Payment
Click on
Banner
Product
View
1.0 2/3
1/3
1/2
1/4
Drop out
3/4
1/4
1
1/4
User next action prediction
12. Contextual Partitioning
• Approaches to create local models:
o Horizontal partitions
Users
from
Europe
Users
from
South
America
Session 1 Search Refine Search Click on
Banner
Product
View
Payment
Session 3 Product
View
Payment
Session 3 Search Refine Search Refine Search Click on
Banner
Session 4 Search Refine Search Click on
Banner
Product
View
Payment
Session 5 Product
View
Click on
Banner
Search
13. Contextual Partitioning
• Approaches to create local models:
o Horizontal partition
o Vertical partition :
• Two types of behavior:
o Ready to by – (Product View, Payment)
o Just browsing – (Search, Refine Search, Click on Banner)
Session 1 Search Refine Search Click on
Banner
Product
View
Payment
Session 3 Product
View
Payment
Session 3 Search Refine Search Refine Search Click on
Banner
Session 4 Search Refine Search Click on
Banner
Product
View
Payment
Session 5 Product
View
Click on
Banner
Search
14. Contextual Partitioning
• Approaches to create local models:
o Horizontal partition
o Vertical partition :
• Two types of behavior:
o Ready to by – (Product View, Payment)
o Just browsing – (Search, Refine Search, Click on Banner)
Session 1 Search Refine Search Click on
Banner
Product
View
Payment
Session 3 Product
View
Payment
Session 3 Search Refine Search Refine Search Click on
Banner
Session 4 Search Refine Search Click on
Banner
Product
View
Payment
Session 5 Product
View
Click on
Banner
Search
15. Context Definition
• Intuition about Context: change of user intents
o User is looking for the product
o User is ready to buy
Search
Refine
Search
Payment
Click on
Banner
Product
View
Intent: looking for product Intent: ready to buy
16. Context Discovery
• Context definition: change of user intents
o User is looking for the product
o User is ready to buy
• Context discovery – apply hierarchical clustering in
order to maximize prediction accuracy
Search
Refine
Search
PaymentClick
Product
View
Intent: looking for product Intent: ready to buy
19. Context Integration
Example
……………
…
C1 C2
C3 Cn
Contextual
features Fs
DATA Environment
L1 L2 L3
Lk
Contextual Categories
Individual Learners
Mapping G
Mapping H
Context
Discovery
20. Thank you!
• Context identification and
integration it into prediction models
• Accurately predicting users’ desired
actions and understanding
behavioral patterns of users in
various web-applications
• Personalization and adaptation to
diverse customer need and
preferences
• Accounting for the practical needs
within the considered application
areas.
21. Summary
• The main goal is to develop a generic framework
for context-aware systems for Web Predictive
Analytics
• In order to archive this goal we need to answer the
following questions:
o How to define the context in predictive web
analytics?
o How to connect context with the prediction
process in predictive web analytics?
Questions?
23. Context-aware systems
Context definition
Context Integration Method
Application
Context-aware system
Recommendation
systems
Computational
Advertisement
Information
Retrieval
Normalization
Expansion
Classifier Selection
Classification Adjustment
Weighting
Domain Expert
Clustering
Contextual feature identification
24. History of context
definition and discovery
Context Year
Location 1992
Taxonomy of explicit context 1999
Predictive features vs. contextual 2002
Hidden context: (clustering, mixture
models)
2004
Contextual bandits 2007
History of previous interaction 2008
Independence of predicted class 2011
Two level prediction model 2012
Focus on Context Discovery 2012 -
Timeline
25. Research Goal
• Our research aims to develop a generic framework
and corresponding techniques for introducing the
contextual information in Predictive Web Analytics
and accounting for the practical needs within the
considered application areas.
Notes de l'éditeur
Good day! My name is …, I am from Eindhoven university of Technology, My supervisor is MykolaPechenizkiy.I am happy present.
Let’s define what WPA is. Def on slideApplication of WPA on slide, make example – personalized search result page
Let’s consider a simple design of web predictive analytics.In WPA we commonly use the following formulation: ON SLIDE to solve WPA tasks.
Let’s consider concrete example of user’ next action prediction.How we can predict next user’ action?We have user’ web log. The most commonly used model for sequence prediction is Markov Models.
We can process our historical data and build a user navigation graph. Which is suitable to train a markov models.
Let’s consider what is a context. ON THE SLIDE
Many taxonomies were built for explictit context.Physical EnvironmentFactors -> Conditions -> Weather -> Context can have different grannularity
The question what kind of context could be useful for WPA and how we can utilize it.Our research aims to develop a generic framework and corresponding techniques for introducing the contextual information in Predictive Web Analytics and accounting for the practical needs within the considered application areas.
Main Resear question on the slideMain focus of my work is to define and discovery useful context and find ways to integrate this context.
We can process our historical data and build a user navigation graph. Which is suitable to train a markov models.We can incorporate context to build local markov models which is easy to train and parametrized.
Let’s consider the following exampleExplain how it can be helpful
In order to build a local Markov models we can partiotion our dat.There are two types of partitioning: horizontal and vertical. Let’s start from horizontal. As context we use “user geo location”.
Vertical partition based on user intent
We have defined a context as “user intent”, which help us to make a vertical partition.
How we can discovery user intent?
My focus is building local models.Contextual information Input adjustment, model selection, output adjustment
Let’s consider our running example for model selection.
More general schema
Addчто сделано
Лучше убрать
Add examples instead of historyRunning examplesAdd definition of explicit and implicit contexts
Our research aims to develop a generic framework and corresponding techniques for introducing the contextual information in Predictive Web Analytics and accounting for the practical needs within the considered application areas