Life events have a deep effect on the individual’s spending habits and purchase patterns. Knowledge on life events opens up opportunities for up- and cross-selling. Can we detect and predict the occurrence of life events based on social media user activities?
7. “A major event in a personal life that will
trigger a process of reconsidering current
behaviour”
(van der Waerden et al, 2003)
8. Formalising life
events
based on Holmes-Rahe stress scale
1. Death of spouse (100)
2. Divorce (73)
3. Marital separation (65)
4. Jail term (63)
5. Death of close family member
(63)
6. Personal injury or illness (53)
7. Marriage (50)
8. Fired at work (47)
9. Marital reconciliation (45)
10. Retirement (45)
10. Life events have a deep effect on the individual’s
spending habits and purchase patterns
Knowledge on life events opens up opportunities
for up- and cross-selling
11. How to do it?
Explicit
surveys, interviews
Implicit
monitoring behaviour
12. Detect and predict the occurrence of
life events based on social media
user activities
Our goal:
14. Feasibility - The Tech side
Model-based AI-based
+
socio-demographics
context
social media actions
15. Model-based
• You build a model of how users react to
the occurrence of an event on social
media
• Join new groups on Facebook?
• Follow new accounts on Twitter?
• Post help requests?
• Pictures on Instagram?
16. AI-based
• You create an annotated training set
(timeline/occurrence of life event)
• Let ML do the job
17. AI-based (under the hood)
Raw content
(text, images,
hashtags,
groups etc.)
Enriched
content
Entity extraction,
categorisation,
sentiment, automated
image annotation
Relevance
(with accuracy)
Classifier
18. AI-based - looking for changes
• Looking for behavioural changes
—> Looking for changes in
activity patterns
• Analysing the frequency of posts
focussed on a given life events
(and the reaction of the user’s
social network)
0
17.5
35
52.5
70
Jan Mar May Jul Sep Nov
19. AI-based: Issues
• Diversity
• Different people, different age groups, different cultures etc. behave wildly different when it
comes to sharing life events related information
• Different behaviour on different social media platforms
• Holds also for model-based approaches
• Data
• For training classifiers
• For benchmarking purposes
• For performance evaluation
21. Feasibility - The legal side
Is it legal?
What about GDPR?
Is it ethical?
22. GDPR?
• We are dealing with the processing of personal information (Art. 4
GDPR)
• The user must express explicit consent (Art. 5 GDPR)
• This does not fall within special categories of personal data (Art. 9
GDPR)
• We use personal information for automated, individual decision-
making, including profiling (Art. 22 GDPR)
24. Summary
• Life event detection and prediction from social media activities is
feasible
• It can comply with GDPR, but informed consent to be handled with
care
• Science&Tech: Missing reference dataset and benchmarks
• Innovation&Business: Missing empirical validation of value for
marketers
To know more: daniele.miorandi@u-hopper.com
25. Acknowledgement
This project has received funding from the European Union’s
Horizon 2020 research and innovation programme under grant
agreement No 739783
The information and views set out in this presentation are those
of the author(s) and do not necessarily reflect the official
opinion of the European Union. Neither the European Union
institutions and bodies nor any person acting on their behalf
may be held responsible for the use which may be made of the
information contained therein.