4. 4
100 million people live in Latin
America without access to the
internet
because of…
geography, orography, economic
conditions, density y population
dispersion, …
…current technologies are
insuficient
Big data & Connect the Unconnected
We are using Big Data, Machine
Learning and AI to :
• Localize and identify the
unconnected,
• Optimize our transport
network
• Optimize our network
operations
5. 5
With AI we pinpoint the
actual demand…
+
Data: satellite images
model: Neural network
Training: census data
Result: actual population
distribution map
…and we compare it with the network
map…
=
Data: geolocalized mobile
sessions
Model: coverage polygons
Training: Telefónica coverage and
regulatory data
Result: Real coverage map
… we then deduce the unconnected
population
Result: clasification y
clusterization of the
uncovered population
Where are the unconnected?
6. 6
2o
Neural network literally
identify every building
3o
4o
1o Super HD digital satellite
images
Houses are grouped and
compared with census
data
RESULT: HD population
distribution model
Of the population localized with
<3% false positives95 %
With AI we pinpoint the actual demand…
7. 7
1o
2o
3o
4o
Geolocalized mobile
sesión and network data
Towers location and
estimation of the
coverage
Calibration based on
reported coverage and
Telefonica data
RESULT: actual network
coverage
Is the median error for tower
location240m
…and we compare it with the network map…
8. 8
Transport networks are the most expensive part of deploying
connectivity to remote areas. Optimizing transport route has a huge
impact on the sustainability of a network.
Big data and graph theory enable us to analyze and compute the
optimal transport deployments.
AI to optimize the transport network
9. 9
We integrate the infrastructure
information…
…we then generate weighted graphs
projecting the opportunity..
… we then compute the optimal
deployment
+ =
Data: detailed infrastructure
Model: geolocalized data
Result: infratructure map of twoers
and roads
Data: clustered opportunity
Model: graph generation
Result: Graph with population
weighted transport options
Model: Graph analysis (Shortest
path, Steiner tree)
Result: population cluster optimized
transport routes
AI to optimize the transport network
10. 10
To connect very remote zones, optimizing operations and minimizing
maintenance and upgrade is key to a sustainable operational model.
AI to optimize networks operation
11. 11
We identify the cells that have
the most probability of failure
Automated preventative error
prediction
We optimize and reduce
network operations
+ =
Data: network metrics
Model: neural network
Training: historical failure
analysis
Result: possible cells failure
map
Data: alarms and outages
Model: task automation
Result: automated supervisión
and preventative outage
prediction
Result: supervisión is
automated and less incidence
of failure
AI to optimize networks operation
12. 12
We capture in real time health
data of network elements
1o 2o
We determine the conditions
of the most common failures We predict which
cells could fail
3o
We create a system that
monitors alarms and evaluates
the scenarios
4o
Supervision is
automated (Operations
Bot) and we predict
failures
AI to identify and predict network outages
13. 13
So what does all of this has to
do with climate change and
natural disasters?
14. 14
El nino costero 2016 - Peru
http://www.conclusion.com.ar/internacionales/peru-la-mitad-de-la-poblacion-vive-en-emergencia-por-el-fenomeno-del-nino-costero/03/2017/ https://elcomercio.pe/peru/nino-costero-549-puentes-6-mil-km-vias-afectadas-416020