SlideShare a Scribd company logo
1 of 17
Download to read offline
Impact of natural hazard on consumer behavior:
The case of the 2017 El Niño phenomenon (ENSO) in Peru
Vincent Gauthier • Telecom SudParis - Institut Polytechnique de Paris
Network Science for the Sustainable Development Goals
NetSci 2020 Satellite Symposium • Sept 20th, 2020
Slides.https://bit.ly/355pp4H
! @vincentgauthier
Joint work with
Hugo Alatrista-Salas, Miguel Nuñez del Prado Cortez (Univ. Pacifico, Peru)
Monique Becker (Telecom SudParis - Institut Polytechnique de Paris, France)
Background information (I)
Peru.
Peru is a representative democratic republic divided into 25
regions.
1/3 of the peruvians live in the capital city: Lima.
Peru is a developing country, ranking 82nd on the Human
Development Index, with a high level of human development.
Since 2000, Peru is one of the region's most prosperous
economies with an average growth rate of 5.9% and it has
one of the world's fastest industrial growth rates at an
average of 9.6%.
Source: Wikipedia
The 2017 ENSO Events.
The capital city of Lima has been impacted by the 2017 El Niño
costero (coastal) events.
Two mains events: heavy rains and flooding in regions near the
Pacific coast of Peru around mid February and at the end of March
2017.
The aftermath of the two main events on the population have
been mainly due to flooding.
Source: EU Emergency Response Coordination Centre (ERCC)
2
Background information (II)
Understanding the El Niño Costero of 2017: The Definition Problem and Challenges of Climate Forecasting and Disaster Responses.
Ramírez, I. J. & Briones, F.
Int. J. Disaster Risk Sci. (2017).
El Niño–Southern Oscillation (ENSO).
The Humboldt current is a cold stream from antarctica that
flows along the coast of Peru. The Humboldt current has a
cooling influence on the climate of Chile, Peru, etc.
The ENSO is the cycle of warm and cold sea surface
temperature (SST) of the tropical central and eastern Pacific
Ocean.
Impacts.
The ENSO events appear on average every 4 years in Peru at
about Christmas time. This warming of the sea temperature
causes a shift in the atmospheric circulation and it brings
heavy rainfall and tropical cyclone formation increases over
the tropical Pacific Ocean.
In 2017, took place a phenomena called El Niño costero (coastal)
which is a different type of El Niño: This phenomena was
composed of two events that took place between January and
April 2017 that had a large impact on the population of Peru.
Some key figures (OCHA):
§ 1 million affected individuals
§ 639 affected bridges
§ 8,481 km of affected roads
§ 230,317 damaged houses
§ 5,244 ha of deteriorated crops
Source:NOAAhttps://bit.ly/3l6jx0E
Source: Wikipedia
Source:https://bit.ly/2EkWZZ9
3
Motivations
In this study we aim to develop a set of tools to measure socio-economic characteristic response of a
population in the aftermath of natural hazard.
1. We think that consumer behavior is a key variable to understand people needs during a period of climatic
stress.
2. The study of consumer behavior enables fine-grained analysis of the population behavior.
3. There is a growing literature that tries to understand what is the level resilience of the supply chain and
the retail business during disaster events.
4. We focused our attention on the greater area of the capital city of Peru, Lima.
Can we characterize the societal response of a population during and after a natural hazard with a
consumption dataset ?
Carleton TA, Hsiang SM. Social and economic impacts of climate. Science, aad9837, (2016). doi:10.1126/science.aad9837.
Strengthening Post-Hurricane Supply Chain Resilience. National Academies Press, (2020).
“For centuries,thinkers have considered whether and how climatic conditions—such as temperature,rainfall,and violent
storms—influence the nature of societies and the performance of economies.“
--Carton et al
4
Bank transaction dataset
Dataset description.
From June 2016 to May 2017, with approximately 1.5 M clients, 55,000 distinct
merchants, and 116.8 M transactions from both credit and debit cards from BBVA.
Features describing the clients such as anonymous ID, age, gender.
Features describing the transaction, like the timestamp, amount spent in monetary
unit.
Features associated with the bank agency, namely the region, province, and
district, in which the agency of the client was located.
Features characterizing the merchants, such as merchant ID, merchant name,
merchant address, the MCC and the coordinates of the merchants.
MCC: Merchant Category Classification
Transactions distribution.
5
1 Grocery Stores and Supermarkets 5 Betting
2 department stores 6 Universities and professional schools
3 Eating places and restaurants 7 Drug stores and Pharmacies
4 Service stations 8 Tax payments
Top 8 categories in our dataset.
Methods and results
I. Macroscopic approach
a) Events detection with the Kullback-Leibler divergence
b) Causal Impacts
II. Microscopic approach
a) Mobility Markov Chain (MMC).
b) Business ranking evolution in time (PageRank)
c) Core networks evolution of the transaction graph
III. Conclusion
Macroscopic approach
Events detection with the Kullback-Leibler
divergence.
We regroup the different purchases into a set of
relevant categories.
We compute de distribution of each category at a
given time and compare it with the distribution at the
next time instant or the mean distribution.
If the two distributions are similar, the KLD of the two
distributions will be close to zero.
Kullback-Leibler divergence.
7
Macroscopic approach
Causal Impact.
Given a specific time, the Causal
impact captures causality by
measuring the difference between
two different time series: one
series under treatment, and
another series not under
treatment.
Timeseries: the transaction volume
of the different districts of the
greater area of Lima
Brodersen KH, Gallusser F, Koehler J, Remy N, Scott SL, et al. Inferring causal impact using
Bayesian structural time-series models. The Annals of Applied Statistics. (2015). 8
Microscopic approach
Mobility Markov Chain (MMC).
We measure the changes in people
behavior using the MMC to quantify
changes in both individual purchase
category and frequent whereabouts.
MMC Method.
Average NDCG in each of the regions of the
greater area of Lima.
9
We compute mean behavior (for all clients)
in each of the regions of the greater area of Lima
Microscopic approach
PageRank of the transaction graph.
1. Based on the analysis of the
purchases made by clients we
build the transaction graph.
2. We slice the transaction
graph daily.
3. We quantify how businesses
reacted during the event by
studying the evolution of the
PageRank of each merchant
in the transaction graph.
Fig 2. Examples.
Merchant B
merchant A
merchant C
Merchant D
client i
client j
client j
client i
client j
client
client i
client j
Sequence of pruchase
between [t1
, t8
[
Sequence of pruchase
between [t8
, t16
[
(t1
, A), (t3
, B), (t7
, C)
(t2
, C), (t4
, D), (t5
, A), (t7
, B)
...
...
a
b
w=2
w=1 w=1
w=1
A
B
C
D
c
Fig 1. Transaction Graph.
10
Microscopic approach
Time series clustering.
We cluster the merchants’ ranking
evolution in order to observe what was
the merchants’ behavior during the main
ENSO events.
We found six main behaviors:
1. C#2 and C#5: transient period during
the main event in February.
2. C#1 and C#0: long term changes in the
ranking.
3. C#3: no noticeable ranking evolution
during the events.
4. C#4: ranking evolution during the
second event.
−1
0
1
cluster #0
−1
0
1
cluster #3
−1
0
1
cluster #1
−1
0
1
cluster #4
01-08
01-28
02-17
03-09
03-29
04-18
−1
0
1
cluster #2
01-08
01-28
02-17
03-09
03-29
04-18
−1
0
1
cluster #5
Clusters.
ML Framework.
11
−50
0
50
∆error(%)
Cluster #0 Cluster #1 Cluster #2
10−3
10−2
10−1
100
−50
0
50
|∆error|(%)
Cluster #3 Cluster #4 Cluster #5
Food
Unlabeled
Health
Clothing
Gas
Technology
Trans.
Housing
Insurance
Hotels
NightLife
10−3
10−2
10−1
100
Food
Gas
Health
Unlabeled
Clothing
Hotels
NightLife
Technology
Trans.
Food
Gas
Health
Unlabeled
Clothing
NightLife
Technology
Trans.
Hotels
Housing
a
100
)
Cluster #0 Cluster #1 Cluster #2
b
Microscopic approach
Interpretation.
Summary of the most important
business’s categories found in each
cluster.
For instance, some gas stations have
been out of order (#5) when others
have been over solicited (#2).
Distribution of the business's categories per cluster. The top of each
figure indicates the relative differences between the proportion of a
given category inside a cluster and the proportion of that category in
our dataset.
Fig 8. Time series clustering.
Table 1. Summary of most important merchant categories found in each cluster
cluster ranking over represented categories under represented categories
#0 0
health food
#1 0
gas, clothing unlabeled stores
#2
0
health, gas, technology, trans-
ports
food
#3 0
health food
#4 0
food, gas health, clothing
#5 0
gas, food clothing, night-life
In Fig. 9, two additional phenomena can be observed: first, an increase in insurance
purchases after the first event in February (cluster 3), and second, an increase in
purchases of technology related items during first event (cluster 2). We observe a surge
in the purchased of new insurance policies after the first event. It should be noted that
Peru is an underinsured country (only approximately three in hundred houses are
insured), and in the aftermath of the event it appears that people decided to purchase
new insurance policies. With respect to technology-related items, a query to the
database at our disposal revealed that no transaction were made in that category from
February 15 to 19, 2017. We observed a shift in purchases after February 19. This
behavior appear to be due to the purchases of prepaid cell phone plans.
We note that people’s response to the crisis were very heterogeneous in time, space
12
Microscopic approach
Resilience of the transaction graph : tracking the
evolution of the core/periphery structure of the
transaction graph.
We studied the resilience of the
transaction graph by measuring the
evolution of the size of the core networks
structure.
What is the set of businesses that remain
in the core structure during the events.
Result:
1. The size of the core structure shrinks
during the events
2. The ENSO events are clear anomalies
3. The businesses categories that
remain in the core are food,gas and
health.
0 5 10 15 20 25 30
Kshell
102
103
104
105
#transactions
20/02
22/02
24/03
26/03
a b
Fig 2.Evolution of the size of the core networks structure in time.
Ma A,Mondrag ́on RJ.Rich-Cores in Networks.PLOS ONE.2015;10(3):e0119678. doi:10.1371/journal.pone.0119678.
Fig 1. Evolution of the k-shell structure.
13
Conclusions
Consumption behavior seems to be a good index to characterize the "societal response"
of a population facing a crisis such as a natural hazard.
Coarse-grained analysis reveals the presence of the El Niño Costero phenomenon
and the recovery time in each territories.
Fine-grained analysis demonstrates a change in individuals' purchasing patterns and
in merchant relevance because of the climatic event. Also different types of
businesses have been impacted differently during the events. There is a wide range
of patterns both in space and time (duration).
Overall, despite a clear slowdown of the economic activities during the natural
hazard, the results indicate that society withstood the stress given that the country
is accustomed to this type of event.
14
Perspectives
Measuring the societal response during and after a crisis is becoming more and more
relevant and feasible (the current COVID-19 epidemic is another example).
The societal response is clearly a multimodal phenomena. One modality cannot
encompass all the facets of the problem.
Different crisis may trigger different societal responses ? is there a universal behavior?
On the contrary, one type of crisis may have regional specificity ?
Multiscale approach need to be developed: long-term impact and short-term shift in
behavior need to be studied more deeply.
15
Thank !!!!!
Impact of natural hazard on consumer behavior:
The case of the 2017 El Nino phenomenon
(ENSO) in Peru
Impact of natural hazard on consumer behavior: The
case of the 2017 El Nino phenomenon (ENSO) in Peru.
Hugo Alatrista-Salas, Vincent Gauthier, Miguel Nuñez
del Prado Cortez, Monique Becker, 2020.
Dataset. https://doi.org/10.7910/DVN/LYXBGR
arXiv.2008.04887
! @vincentgauthier
" https://complex.luxbulb.org/
# vincent.gauthier@telecom-sudparis.eu
Dataset Demography
0 0 0 2 0 4 0 6 0 8 1 0
f
0 00
0 25
0 50
0 75
1 00
C(f)
class 9
class 8
class 7
class 6
class 5
class 4
class 3
class 2
class 1
1 2 3 4 5 6 7 8 9
Class
0
100
200
300
Size(103
)
0
5
10
(103
)
Size
<P>
<P>
35
40
45
averageage
class
0.25
0.50
0.75
Fractionofwomen
b
c d
a
e
02500050000
Men (pop)
0 25000 50000
Women (pop)
20-24
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
70-74
75-79
80-84
85-89
90-94
age 9: rich
8: middle upper
7: lower upper
6: upper middle
5: middle
4: lower middle
3: lower
2: middle lower
1: poor
17

More Related Content

Similar to Network Science for the Sustainable Development Goals

Jorge Katz - Seminario 'Nuevos enfoques sobre políticas de innovación'
Jorge Katz - Seminario 'Nuevos enfoques sobre políticas de innovación'Jorge Katz - Seminario 'Nuevos enfoques sobre políticas de innovación'
Jorge Katz - Seminario 'Nuevos enfoques sobre políticas de innovación'Fundación Ramón Areces
 
125936hdbvjbdscjsdcnndsvjscndjjcdn462.pdf
125936hdbvjbdscjsdcnndsvjscndjjcdn462.pdf125936hdbvjbdscjsdcnndsvjscndjjcdn462.pdf
125936hdbvjbdscjsdcnndsvjscndjjcdn462.pdfNhuQuynh241093
 
1259364djvbdjvndjvndvdsjsssjsnkfnv62.pdf
1259364djvbdjvndjvndvdsjsssjsnkfnv62.pdf1259364djvbdjvndjvndvdsjsssjsnkfnv62.pdf
1259364djvbdjvndjvndvdsjsssjsnkfnv62.pdfNhuQuynh241093
 
Marketing Management, 14Chapter 3 Collecting Information and Fo.docx
Marketing Management, 14Chapter 3 Collecting Information and Fo.docxMarketing Management, 14Chapter 3 Collecting Information and Fo.docx
Marketing Management, 14Chapter 3 Collecting Information and Fo.docxinfantsuk
 
Gas Laws Essay. Online assignment writing service.
Gas Laws Essay. Online assignment writing service.Gas Laws Essay. Online assignment writing service.
Gas Laws Essay. Online assignment writing service.Tanya Collins
 
Horizon Scan: ICT and the Future of Retail
Horizon Scan: ICT and the Future of RetailHorizon Scan: ICT and the Future of Retail
Horizon Scan: ICT and the Future of RetailEricsson
 
Cyber Insurance as Digital Strategy
Cyber Insurance as Digital StrategyCyber Insurance as Digital Strategy
Cyber Insurance as Digital StrategyRandeep Sudan
 
Twitter Based Sentimental Analysis of Impact of COVID-19 on Economy using Naï...
Twitter Based Sentimental Analysis of Impact of COVID-19 on Economy using Naï...Twitter Based Sentimental Analysis of Impact of COVID-19 on Economy using Naï...
Twitter Based Sentimental Analysis of Impact of COVID-19 on Economy using Naï...CSCJournals
 
Essay About Medicine And Ethics
Essay About Medicine And EthicsEssay About Medicine And Ethics
Essay About Medicine And EthicsLydia Jana
 
The impact of the FinTech revolution on the future of banking.pdf
The impact of the FinTech revolution on the future of banking.pdfThe impact of the FinTech revolution on the future of banking.pdf
The impact of the FinTech revolution on the future of banking.pdfTODICHIRALIA
 

Similar to Network Science for the Sustainable Development Goals (12)

Jorge Katz - Seminario 'Nuevos enfoques sobre políticas de innovación'
Jorge Katz - Seminario 'Nuevos enfoques sobre políticas de innovación'Jorge Katz - Seminario 'Nuevos enfoques sobre políticas de innovación'
Jorge Katz - Seminario 'Nuevos enfoques sobre políticas de innovación'
 
125936hdbvjbdscjsdcnndsvjscndjjcdn462.pdf
125936hdbvjbdscjsdcnndsvjscndjjcdn462.pdf125936hdbvjbdscjsdcnndsvjscndjjcdn462.pdf
125936hdbvjbdscjsdcnndsvjscndjjcdn462.pdf
 
1259364djvbdjvndjvndvdsjsssjsnkfnv62.pdf
1259364djvbdjvndjvndvdsjsssjsnkfnv62.pdf1259364djvbdjvndjvndvdsjsssjsnkfnv62.pdf
1259364djvbdjvndjvndvdsjsssjsnkfnv62.pdf
 
Lesson 5
Lesson 5Lesson 5
Lesson 5
 
Marketing Management, 14Chapter 3 Collecting Information and Fo.docx
Marketing Management, 14Chapter 3 Collecting Information and Fo.docxMarketing Management, 14Chapter 3 Collecting Information and Fo.docx
Marketing Management, 14Chapter 3 Collecting Information and Fo.docx
 
Gas Laws Essay. Online assignment writing service.
Gas Laws Essay. Online assignment writing service.Gas Laws Essay. Online assignment writing service.
Gas Laws Essay. Online assignment writing service.
 
Horizon Scan: ICT and the Future of Retail
Horizon Scan: ICT and the Future of RetailHorizon Scan: ICT and the Future of Retail
Horizon Scan: ICT and the Future of Retail
 
Cyber Insurance as Digital Strategy
Cyber Insurance as Digital StrategyCyber Insurance as Digital Strategy
Cyber Insurance as Digital Strategy
 
Twitter Based Sentimental Analysis of Impact of COVID-19 on Economy using Naï...
Twitter Based Sentimental Analysis of Impact of COVID-19 on Economy using Naï...Twitter Based Sentimental Analysis of Impact of COVID-19 on Economy using Naï...
Twitter Based Sentimental Analysis of Impact of COVID-19 on Economy using Naï...
 
Risk Management in Financial Innovations and Sustainable Development in Nigeria
Risk Management in Financial Innovations and Sustainable Development in NigeriaRisk Management in Financial Innovations and Sustainable Development in Nigeria
Risk Management in Financial Innovations and Sustainable Development in Nigeria
 
Essay About Medicine And Ethics
Essay About Medicine And EthicsEssay About Medicine And Ethics
Essay About Medicine And Ethics
 
The impact of the FinTech revolution on the future of banking.pdf
The impact of the FinTech revolution on the future of banking.pdfThe impact of the FinTech revolution on the future of banking.pdf
The impact of the FinTech revolution on the future of banking.pdf
 

Recently uploaded

Data Science Jobs and Salaries Analysis.pptx
Data Science Jobs and Salaries Analysis.pptxData Science Jobs and Salaries Analysis.pptx
Data Science Jobs and Salaries Analysis.pptxFurkanTasci3
 
代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改
代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改
代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改atducpo
 
Data Warehouse , Data Cube Computation
Data Warehouse   , Data Cube ComputationData Warehouse   , Data Cube Computation
Data Warehouse , Data Cube Computationsit20ad004
 
Delhi Call Girls CP 9711199171 ☎✔👌✔ Whatsapp Hard And Sexy Vip Call
Delhi Call Girls CP 9711199171 ☎✔👌✔ Whatsapp Hard And Sexy Vip CallDelhi Call Girls CP 9711199171 ☎✔👌✔ Whatsapp Hard And Sexy Vip Call
Delhi Call Girls CP 9711199171 ☎✔👌✔ Whatsapp Hard And Sexy Vip Callshivangimorya083
 
RadioAdProWritingCinderellabyButleri.pdf
RadioAdProWritingCinderellabyButleri.pdfRadioAdProWritingCinderellabyButleri.pdf
RadioAdProWritingCinderellabyButleri.pdfgstagge
 
Brighton SEO | April 2024 | Data Storytelling
Brighton SEO | April 2024 | Data StorytellingBrighton SEO | April 2024 | Data Storytelling
Brighton SEO | April 2024 | Data StorytellingNeil Barnes
 
04242024_CCC TUG_Joins and Relationships
04242024_CCC TUG_Joins and Relationships04242024_CCC TUG_Joins and Relationships
04242024_CCC TUG_Joins and Relationshipsccctableauusergroup
 
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...Suhani Kapoor
 
From idea to production in a day – Leveraging Azure ML and Streamlit to build...
From idea to production in a day – Leveraging Azure ML and Streamlit to build...From idea to production in a day – Leveraging Azure ML and Streamlit to build...
From idea to production in a day – Leveraging Azure ML and Streamlit to build...Florian Roscheck
 
VIP High Profile Call Girls Amravati Aarushi 8250192130 Independent Escort Se...
VIP High Profile Call Girls Amravati Aarushi 8250192130 Independent Escort Se...VIP High Profile Call Girls Amravati Aarushi 8250192130 Independent Escort Se...
VIP High Profile Call Girls Amravati Aarushi 8250192130 Independent Escort Se...Suhani Kapoor
 
VIP Call Girls in Amravati Aarohi 8250192130 Independent Escort Service Amravati
VIP Call Girls in Amravati Aarohi 8250192130 Independent Escort Service AmravatiVIP Call Girls in Amravati Aarohi 8250192130 Independent Escort Service Amravati
VIP Call Girls in Amravati Aarohi 8250192130 Independent Escort Service AmravatiSuhani Kapoor
 
VIP High Class Call Girls Bikaner Anushka 8250192130 Independent Escort Servi...
VIP High Class Call Girls Bikaner Anushka 8250192130 Independent Escort Servi...VIP High Class Call Girls Bikaner Anushka 8250192130 Independent Escort Servi...
VIP High Class Call Girls Bikaner Anushka 8250192130 Independent Escort Servi...Suhani Kapoor
 
Market Analysis in the 5 Largest Economic Countries in Southeast Asia.pdf
Market Analysis in the 5 Largest Economic Countries in Southeast Asia.pdfMarket Analysis in the 5 Largest Economic Countries in Southeast Asia.pdf
Market Analysis in the 5 Largest Economic Countries in Southeast Asia.pdfRachmat Ramadhan H
 
Beautiful Sapna Vip Call Girls Hauz Khas 9711199012 Call /Whatsapps
Beautiful Sapna Vip  Call Girls Hauz Khas 9711199012 Call /WhatsappsBeautiful Sapna Vip  Call Girls Hauz Khas 9711199012 Call /Whatsapps
Beautiful Sapna Vip Call Girls Hauz Khas 9711199012 Call /Whatsappssapnasaifi408
 
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130Suhani Kapoor
 
Schema on read is obsolete. Welcome metaprogramming..pdf
Schema on read is obsolete. Welcome metaprogramming..pdfSchema on read is obsolete. Welcome metaprogramming..pdf
Schema on read is obsolete. Welcome metaprogramming..pdfLars Albertsson
 
EMERCE - 2024 - AMSTERDAM - CROSS-PLATFORM TRACKING WITH GOOGLE ANALYTICS.pptx
EMERCE - 2024 - AMSTERDAM - CROSS-PLATFORM  TRACKING WITH GOOGLE ANALYTICS.pptxEMERCE - 2024 - AMSTERDAM - CROSS-PLATFORM  TRACKING WITH GOOGLE ANALYTICS.pptx
EMERCE - 2024 - AMSTERDAM - CROSS-PLATFORM TRACKING WITH GOOGLE ANALYTICS.pptxthyngster
 

Recently uploaded (20)

Data Science Jobs and Salaries Analysis.pptx
Data Science Jobs and Salaries Analysis.pptxData Science Jobs and Salaries Analysis.pptx
Data Science Jobs and Salaries Analysis.pptx
 
代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改
代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改
代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改
 
Data Warehouse , Data Cube Computation
Data Warehouse   , Data Cube ComputationData Warehouse   , Data Cube Computation
Data Warehouse , Data Cube Computation
 
Delhi Call Girls CP 9711199171 ☎✔👌✔ Whatsapp Hard And Sexy Vip Call
Delhi Call Girls CP 9711199171 ☎✔👌✔ Whatsapp Hard And Sexy Vip CallDelhi Call Girls CP 9711199171 ☎✔👌✔ Whatsapp Hard And Sexy Vip Call
Delhi Call Girls CP 9711199171 ☎✔👌✔ Whatsapp Hard And Sexy Vip Call
 
RadioAdProWritingCinderellabyButleri.pdf
RadioAdProWritingCinderellabyButleri.pdfRadioAdProWritingCinderellabyButleri.pdf
RadioAdProWritingCinderellabyButleri.pdf
 
Brighton SEO | April 2024 | Data Storytelling
Brighton SEO | April 2024 | Data StorytellingBrighton SEO | April 2024 | Data Storytelling
Brighton SEO | April 2024 | Data Storytelling
 
04242024_CCC TUG_Joins and Relationships
04242024_CCC TUG_Joins and Relationships04242024_CCC TUG_Joins and Relationships
04242024_CCC TUG_Joins and Relationships
 
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...
 
From idea to production in a day – Leveraging Azure ML and Streamlit to build...
From idea to production in a day – Leveraging Azure ML and Streamlit to build...From idea to production in a day – Leveraging Azure ML and Streamlit to build...
From idea to production in a day – Leveraging Azure ML and Streamlit to build...
 
VIP High Profile Call Girls Amravati Aarushi 8250192130 Independent Escort Se...
VIP High Profile Call Girls Amravati Aarushi 8250192130 Independent Escort Se...VIP High Profile Call Girls Amravati Aarushi 8250192130 Independent Escort Se...
VIP High Profile Call Girls Amravati Aarushi 8250192130 Independent Escort Se...
 
Russian Call Girls Dwarka Sector 15 💓 Delhi 9999965857 @Sabina Modi VVIP MODE...
Russian Call Girls Dwarka Sector 15 💓 Delhi 9999965857 @Sabina Modi VVIP MODE...Russian Call Girls Dwarka Sector 15 💓 Delhi 9999965857 @Sabina Modi VVIP MODE...
Russian Call Girls Dwarka Sector 15 💓 Delhi 9999965857 @Sabina Modi VVIP MODE...
 
VIP Call Girls in Amravati Aarohi 8250192130 Independent Escort Service Amravati
VIP Call Girls in Amravati Aarohi 8250192130 Independent Escort Service AmravatiVIP Call Girls in Amravati Aarohi 8250192130 Independent Escort Service Amravati
VIP Call Girls in Amravati Aarohi 8250192130 Independent Escort Service Amravati
 
VIP High Class Call Girls Bikaner Anushka 8250192130 Independent Escort Servi...
VIP High Class Call Girls Bikaner Anushka 8250192130 Independent Escort Servi...VIP High Class Call Girls Bikaner Anushka 8250192130 Independent Escort Servi...
VIP High Class Call Girls Bikaner Anushka 8250192130 Independent Escort Servi...
 
꧁❤ Aerocity Call Girls Service Aerocity Delhi ❤꧂ 9999965857 ☎️ Hard And Sexy ...
꧁❤ Aerocity Call Girls Service Aerocity Delhi ❤꧂ 9999965857 ☎️ Hard And Sexy ...꧁❤ Aerocity Call Girls Service Aerocity Delhi ❤꧂ 9999965857 ☎️ Hard And Sexy ...
꧁❤ Aerocity Call Girls Service Aerocity Delhi ❤꧂ 9999965857 ☎️ Hard And Sexy ...
 
Deep Generative Learning for All - The Gen AI Hype (Spring 2024)
Deep Generative Learning for All - The Gen AI Hype (Spring 2024)Deep Generative Learning for All - The Gen AI Hype (Spring 2024)
Deep Generative Learning for All - The Gen AI Hype (Spring 2024)
 
Market Analysis in the 5 Largest Economic Countries in Southeast Asia.pdf
Market Analysis in the 5 Largest Economic Countries in Southeast Asia.pdfMarket Analysis in the 5 Largest Economic Countries in Southeast Asia.pdf
Market Analysis in the 5 Largest Economic Countries in Southeast Asia.pdf
 
Beautiful Sapna Vip Call Girls Hauz Khas 9711199012 Call /Whatsapps
Beautiful Sapna Vip  Call Girls Hauz Khas 9711199012 Call /WhatsappsBeautiful Sapna Vip  Call Girls Hauz Khas 9711199012 Call /Whatsapps
Beautiful Sapna Vip Call Girls Hauz Khas 9711199012 Call /Whatsapps
 
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
 
Schema on read is obsolete. Welcome metaprogramming..pdf
Schema on read is obsolete. Welcome metaprogramming..pdfSchema on read is obsolete. Welcome metaprogramming..pdf
Schema on read is obsolete. Welcome metaprogramming..pdf
 
EMERCE - 2024 - AMSTERDAM - CROSS-PLATFORM TRACKING WITH GOOGLE ANALYTICS.pptx
EMERCE - 2024 - AMSTERDAM - CROSS-PLATFORM  TRACKING WITH GOOGLE ANALYTICS.pptxEMERCE - 2024 - AMSTERDAM - CROSS-PLATFORM  TRACKING WITH GOOGLE ANALYTICS.pptx
EMERCE - 2024 - AMSTERDAM - CROSS-PLATFORM TRACKING WITH GOOGLE ANALYTICS.pptx
 

Network Science for the Sustainable Development Goals

  • 1. Impact of natural hazard on consumer behavior: The case of the 2017 El Niño phenomenon (ENSO) in Peru Vincent Gauthier • Telecom SudParis - Institut Polytechnique de Paris Network Science for the Sustainable Development Goals NetSci 2020 Satellite Symposium • Sept 20th, 2020 Slides.https://bit.ly/355pp4H ! @vincentgauthier Joint work with Hugo Alatrista-Salas, Miguel Nuñez del Prado Cortez (Univ. Pacifico, Peru) Monique Becker (Telecom SudParis - Institut Polytechnique de Paris, France)
  • 2. Background information (I) Peru. Peru is a representative democratic republic divided into 25 regions. 1/3 of the peruvians live in the capital city: Lima. Peru is a developing country, ranking 82nd on the Human Development Index, with a high level of human development. Since 2000, Peru is one of the region's most prosperous economies with an average growth rate of 5.9% and it has one of the world's fastest industrial growth rates at an average of 9.6%. Source: Wikipedia The 2017 ENSO Events. The capital city of Lima has been impacted by the 2017 El Niño costero (coastal) events. Two mains events: heavy rains and flooding in regions near the Pacific coast of Peru around mid February and at the end of March 2017. The aftermath of the two main events on the population have been mainly due to flooding. Source: EU Emergency Response Coordination Centre (ERCC) 2
  • 3. Background information (II) Understanding the El Niño Costero of 2017: The Definition Problem and Challenges of Climate Forecasting and Disaster Responses. Ramírez, I. J. & Briones, F. Int. J. Disaster Risk Sci. (2017). El Niño–Southern Oscillation (ENSO). The Humboldt current is a cold stream from antarctica that flows along the coast of Peru. The Humboldt current has a cooling influence on the climate of Chile, Peru, etc. The ENSO is the cycle of warm and cold sea surface temperature (SST) of the tropical central and eastern Pacific Ocean. Impacts. The ENSO events appear on average every 4 years in Peru at about Christmas time. This warming of the sea temperature causes a shift in the atmospheric circulation and it brings heavy rainfall and tropical cyclone formation increases over the tropical Pacific Ocean. In 2017, took place a phenomena called El Niño costero (coastal) which is a different type of El Niño: This phenomena was composed of two events that took place between January and April 2017 that had a large impact on the population of Peru. Some key figures (OCHA): § 1 million affected individuals § 639 affected bridges § 8,481 km of affected roads § 230,317 damaged houses § 5,244 ha of deteriorated crops Source:NOAAhttps://bit.ly/3l6jx0E Source: Wikipedia Source:https://bit.ly/2EkWZZ9 3
  • 4. Motivations In this study we aim to develop a set of tools to measure socio-economic characteristic response of a population in the aftermath of natural hazard. 1. We think that consumer behavior is a key variable to understand people needs during a period of climatic stress. 2. The study of consumer behavior enables fine-grained analysis of the population behavior. 3. There is a growing literature that tries to understand what is the level resilience of the supply chain and the retail business during disaster events. 4. We focused our attention on the greater area of the capital city of Peru, Lima. Can we characterize the societal response of a population during and after a natural hazard with a consumption dataset ? Carleton TA, Hsiang SM. Social and economic impacts of climate. Science, aad9837, (2016). doi:10.1126/science.aad9837. Strengthening Post-Hurricane Supply Chain Resilience. National Academies Press, (2020). “For centuries,thinkers have considered whether and how climatic conditions—such as temperature,rainfall,and violent storms—influence the nature of societies and the performance of economies.“ --Carton et al 4
  • 5. Bank transaction dataset Dataset description. From June 2016 to May 2017, with approximately 1.5 M clients, 55,000 distinct merchants, and 116.8 M transactions from both credit and debit cards from BBVA. Features describing the clients such as anonymous ID, age, gender. Features describing the transaction, like the timestamp, amount spent in monetary unit. Features associated with the bank agency, namely the region, province, and district, in which the agency of the client was located. Features characterizing the merchants, such as merchant ID, merchant name, merchant address, the MCC and the coordinates of the merchants. MCC: Merchant Category Classification Transactions distribution. 5 1 Grocery Stores and Supermarkets 5 Betting 2 department stores 6 Universities and professional schools 3 Eating places and restaurants 7 Drug stores and Pharmacies 4 Service stations 8 Tax payments Top 8 categories in our dataset.
  • 6. Methods and results I. Macroscopic approach a) Events detection with the Kullback-Leibler divergence b) Causal Impacts II. Microscopic approach a) Mobility Markov Chain (MMC). b) Business ranking evolution in time (PageRank) c) Core networks evolution of the transaction graph III. Conclusion
  • 7. Macroscopic approach Events detection with the Kullback-Leibler divergence. We regroup the different purchases into a set of relevant categories. We compute de distribution of each category at a given time and compare it with the distribution at the next time instant or the mean distribution. If the two distributions are similar, the KLD of the two distributions will be close to zero. Kullback-Leibler divergence. 7
  • 8. Macroscopic approach Causal Impact. Given a specific time, the Causal impact captures causality by measuring the difference between two different time series: one series under treatment, and another series not under treatment. Timeseries: the transaction volume of the different districts of the greater area of Lima Brodersen KH, Gallusser F, Koehler J, Remy N, Scott SL, et al. Inferring causal impact using Bayesian structural time-series models. The Annals of Applied Statistics. (2015). 8
  • 9. Microscopic approach Mobility Markov Chain (MMC). We measure the changes in people behavior using the MMC to quantify changes in both individual purchase category and frequent whereabouts. MMC Method. Average NDCG in each of the regions of the greater area of Lima. 9 We compute mean behavior (for all clients) in each of the regions of the greater area of Lima
  • 10. Microscopic approach PageRank of the transaction graph. 1. Based on the analysis of the purchases made by clients we build the transaction graph. 2. We slice the transaction graph daily. 3. We quantify how businesses reacted during the event by studying the evolution of the PageRank of each merchant in the transaction graph. Fig 2. Examples. Merchant B merchant A merchant C Merchant D client i client j client j client i client j client client i client j Sequence of pruchase between [t1 , t8 [ Sequence of pruchase between [t8 , t16 [ (t1 , A), (t3 , B), (t7 , C) (t2 , C), (t4 , D), (t5 , A), (t7 , B) ... ... a b w=2 w=1 w=1 w=1 A B C D c Fig 1. Transaction Graph. 10
  • 11. Microscopic approach Time series clustering. We cluster the merchants’ ranking evolution in order to observe what was the merchants’ behavior during the main ENSO events. We found six main behaviors: 1. C#2 and C#5: transient period during the main event in February. 2. C#1 and C#0: long term changes in the ranking. 3. C#3: no noticeable ranking evolution during the events. 4. C#4: ranking evolution during the second event. −1 0 1 cluster #0 −1 0 1 cluster #3 −1 0 1 cluster #1 −1 0 1 cluster #4 01-08 01-28 02-17 03-09 03-29 04-18 −1 0 1 cluster #2 01-08 01-28 02-17 03-09 03-29 04-18 −1 0 1 cluster #5 Clusters. ML Framework. 11
  • 12. −50 0 50 ∆error(%) Cluster #0 Cluster #1 Cluster #2 10−3 10−2 10−1 100 −50 0 50 |∆error|(%) Cluster #3 Cluster #4 Cluster #5 Food Unlabeled Health Clothing Gas Technology Trans. Housing Insurance Hotels NightLife 10−3 10−2 10−1 100 Food Gas Health Unlabeled Clothing Hotels NightLife Technology Trans. Food Gas Health Unlabeled Clothing NightLife Technology Trans. Hotels Housing a 100 ) Cluster #0 Cluster #1 Cluster #2 b Microscopic approach Interpretation. Summary of the most important business’s categories found in each cluster. For instance, some gas stations have been out of order (#5) when others have been over solicited (#2). Distribution of the business's categories per cluster. The top of each figure indicates the relative differences between the proportion of a given category inside a cluster and the proportion of that category in our dataset. Fig 8. Time series clustering. Table 1. Summary of most important merchant categories found in each cluster cluster ranking over represented categories under represented categories #0 0 health food #1 0 gas, clothing unlabeled stores #2 0 health, gas, technology, trans- ports food #3 0 health food #4 0 food, gas health, clothing #5 0 gas, food clothing, night-life In Fig. 9, two additional phenomena can be observed: first, an increase in insurance purchases after the first event in February (cluster 3), and second, an increase in purchases of technology related items during first event (cluster 2). We observe a surge in the purchased of new insurance policies after the first event. It should be noted that Peru is an underinsured country (only approximately three in hundred houses are insured), and in the aftermath of the event it appears that people decided to purchase new insurance policies. With respect to technology-related items, a query to the database at our disposal revealed that no transaction were made in that category from February 15 to 19, 2017. We observed a shift in purchases after February 19. This behavior appear to be due to the purchases of prepaid cell phone plans. We note that people’s response to the crisis were very heterogeneous in time, space 12
  • 13. Microscopic approach Resilience of the transaction graph : tracking the evolution of the core/periphery structure of the transaction graph. We studied the resilience of the transaction graph by measuring the evolution of the size of the core networks structure. What is the set of businesses that remain in the core structure during the events. Result: 1. The size of the core structure shrinks during the events 2. The ENSO events are clear anomalies 3. The businesses categories that remain in the core are food,gas and health. 0 5 10 15 20 25 30 Kshell 102 103 104 105 #transactions 20/02 22/02 24/03 26/03 a b Fig 2.Evolution of the size of the core networks structure in time. Ma A,Mondrag ́on RJ.Rich-Cores in Networks.PLOS ONE.2015;10(3):e0119678. doi:10.1371/journal.pone.0119678. Fig 1. Evolution of the k-shell structure. 13
  • 14. Conclusions Consumption behavior seems to be a good index to characterize the "societal response" of a population facing a crisis such as a natural hazard. Coarse-grained analysis reveals the presence of the El Niño Costero phenomenon and the recovery time in each territories. Fine-grained analysis demonstrates a change in individuals' purchasing patterns and in merchant relevance because of the climatic event. Also different types of businesses have been impacted differently during the events. There is a wide range of patterns both in space and time (duration). Overall, despite a clear slowdown of the economic activities during the natural hazard, the results indicate that society withstood the stress given that the country is accustomed to this type of event. 14
  • 15. Perspectives Measuring the societal response during and after a crisis is becoming more and more relevant and feasible (the current COVID-19 epidemic is another example). The societal response is clearly a multimodal phenomena. One modality cannot encompass all the facets of the problem. Different crisis may trigger different societal responses ? is there a universal behavior? On the contrary, one type of crisis may have regional specificity ? Multiscale approach need to be developed: long-term impact and short-term shift in behavior need to be studied more deeply. 15
  • 16. Thank !!!!! Impact of natural hazard on consumer behavior: The case of the 2017 El Nino phenomenon (ENSO) in Peru Impact of natural hazard on consumer behavior: The case of the 2017 El Nino phenomenon (ENSO) in Peru. Hugo Alatrista-Salas, Vincent Gauthier, Miguel Nuñez del Prado Cortez, Monique Becker, 2020. Dataset. https://doi.org/10.7910/DVN/LYXBGR arXiv.2008.04887 ! @vincentgauthier " https://complex.luxbulb.org/ # vincent.gauthier@telecom-sudparis.eu
  • 17. Dataset Demography 0 0 0 2 0 4 0 6 0 8 1 0 f 0 00 0 25 0 50 0 75 1 00 C(f) class 9 class 8 class 7 class 6 class 5 class 4 class 3 class 2 class 1 1 2 3 4 5 6 7 8 9 Class 0 100 200 300 Size(103 ) 0 5 10 (103 ) Size <P> <P> 35 40 45 averageage class 0.25 0.50 0.75 Fractionofwomen b c d a e 02500050000 Men (pop) 0 25000 50000 Women (pop) 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85-89 90-94 age 9: rich 8: middle upper 7: lower upper 6: upper middle 5: middle 4: lower middle 3: lower 2: middle lower 1: poor 17