Vincenzo Sciacca - Insight Cloud Services: costruire ed identificare informazioni attraverso l'analisi e la correlazione dei dati relativi agli individui, eventi, news, IoT, social, open data, dati meteo ecc. Un lavoro in continua evoluzione per selezionare sempre nuove fonti dati e costruire strumenti di analisi cognitivi sempre più efficienti, per comprendere il mondo ed estrarre informazioni, "intuizioni" che ci consentano di prendere decisioni migliori ed innovare velocemente.
2. The Insights Economy
Data is transforming
industries and
professions.
The world
is being reinvented
in code.
In the Cognitive Era
digital intelligence meets
digital business.
Leveraging the new mix
of data
Developers, platforms &
infrastructures
Solutions & Smarter
Industries
Analytics strategic imperatives & their investment axes
Accelerate outcomes &
create deeper business
relevance
Analytics for everyone on
an open, fluid & unified
architecture
Agile integration &
governance of internal,
external & machine-
based data
3. The Insights Economy
Power digital-
physical innovation
Increase operational
efficiency
Manage
risk
Transform customer
engagement
• 100% prediction of aircraft-on-the-
ground events for high-risk engines
• 97% accuracy in predicting engine
events that lead to airline disruption
• 270% increase in cross-sales of
accessory products
• 50% increase in effectiveness of
retention campaigns
• Reduces energy costs by up to 20%
• Saves up to $25M per year keeping
refrigerators at optimal temperatures
• Remote diagnosis of refrigerators to
streamline labor efforts
• 40% increase in identifying suspicious transactions
• 80% increase in productivity
• 200% increase in reporting capabilities
4. Data, Insights, and Action
Data Insight Action
Customer name, address Customer segment Marketing campaigns
+ Customer service requests Propensity to churn, customer
value
Retention action decision
+ Customer transactions Customer value Next best action
+ Twitter Life Event detection Targeted ads and offers
Utility operations + Weather
history + Weather forecast
Energy consumption forecasts Energy production plan
+ Outage history + asset data Outage prediction Crew and resource dispatch
Weather forecast + HC data Asthma risk score Patient alerts, ER staffing
POS data Category Sales performance Pricing actions
+ Weather history Weather adjusted sales
performance
Inventory repositioning
Weather history + Ag yield Variability and Correlation Price Crop Insurance
Equipment sensor data +
weather
Failure prediction Maintenance schedule
Enterprise HR Data Attrition Risk Compensation optimization
Tax payment details Single View of the tax Payer Improve loss prevention
Twitter Where will demonstration
happen
Law enforcement crew dispatch
5. Agricolture Business context
Resources
Users
Supply chain
Supporting
functions
Farmers
Agriculture
Companies
Cooperatives Government R&aD
Inputs
Seeds, Fertilizers
Primary Production
Crops, Farmland Management
Plan Prepare Seed Care Harvest Stock
Natural
Resources
Water, Energy
Environment
Weather, Soil
Equipments
Finance Logistics
Labor
6.
7. The Process to Deliver Actionable Insights
Step 1: Raw Ingredients…Data Acquisition
Harvesting of that raw data separates the usable ingredients from the unusable, cleansing
ensures the data has consistent quality and is contaminant free, and distribution makes
the ingredients available in a convenient location.
The discovery of a new insight takes experimentation with different data ingredients and
the particular mix in which they are used. Once discovered, this recipe is repeatable and
provides value for many applications.
Insights can now be combined into interesting combinations and applied to specific industry
scenarios, improving and optimizing the experience and making them actionable. There is
a wide range of value and application to these insight combinations.
Raw data forms the base ingredients for all insights and actionable processing. The
knowledge of what data is available and where to find it is the first level of value.
Step 2: Harvested & Cleansed Ingredients…Data Cleansing & Curation
Step 3: Ingredients Combined into a Recipe…Insights Processing
Step 4: Recipes Combined to make Meals…Service Delivery
8. Step 1: Finding the right Raw Ingredients
1. Start with Global Contextual Data through
partnerships, purchase, and open data sources:
Weather
Geospatial Boundaries & Points of Interest
Location Aggregation (people & vehicle traffic)
Social Media Network & Content
Demographics
City, State, Federal Government Open Data
Business Registries
Event Calendars
2. Add Specific Commercial & Industry Data
Infrastructure Networks (Utility, Transportation)
Insurance & Risk
Financial Transactions
Customer Relationship Management
Asset Management Systems
Data Acquisition Strategy
9. Step 2: Harvest & Cleanse the Ingredients
1. Catalog and classify data for appropriate usage, understanding and governance
2. Purchase already processed data from selected data providers and partners
3. Statistical profiling and identification of historical patterns and data entry errors
4. Cleanse: correction, elimination of bad data values, sampling for confidence levels
5. Normalize: mapping tables and value translation
6. Manual tool improvements to improve cleansing productivity
7. Selectivity of data values to cleanse, focusing on “just enough integration and cleansing”
Data Cleansing & Curation Strategy
10. Step 3: Combine the Ingredients into a Recipe
Insights Processing Strategy
1. Create incubation areas that blend the harvested and cleansed data combined with a
specific set of customer data.
2. Enable a team of Data Scientists to create analyses and models that can improve specific
line-of-business processes.
3. Create datasets and analytics results that can be used across multiple industries and
applications (weather forecasts can impact retail, government, consumer products,
insurance, life sciences, travel and transportation, and energy and utilities).
4. Create measurement systems that show the effectiveness of the models in the business
process, for continuous improvement.
5. Construct processes to validate and improve insights.
“In the end you should only measure and look at the
numbers that drive action, meaning the data tells you
what you should do next.”
Alex Peiniger CEO, quintly
11. Step 4: Combine Recipes into a Meal
Solutions Delivery Strategy
1. Weave the developed analytics model into an insight that can be run at scale, with easy to
understand and integrate results.
2. Provide those results through standard API libraries and in some cases, User Interfaces
on web or mobile clients.
3. Create a unique customer experience area that enables self-education, self-evident value,
and immediate self-gratification.
“Data are essential, but performance improvements and
competitive advantage arise from analytics models that
allow managers to predict and optimize outcomes.”
Harvard Business Review
12. Network of
190,000+
Weather sensors
Global Lightning
Detection Network
Proprietary Radar
Algorithms
Industry Best
Forecast Modeling
State-of-the-Science
Forecast Technologies
PhD
220+ Full-time
Meteorologists
15-30 day hurricane
forecast
Largest collection of
worldwide forecasts
Proprietary weather
data analytics
A big data and IoT approach to Weather:
The Weather Company delivers 26 billion forecasts per day
4GB of new data
each second
Atmospheric data
from 50,000 flights
per day
2.2 Billion weather
forecast locations Data from 40 Million
mobile devices
13. The Weather Company generates differentiated insights
that can be processed and distributed on a massive scale
Open&Govt
Data
National Weather Service
Weather Stations
High Resolution Radar
Oceanographic Data
TheWeatherCompanyProprietaryand
SourcedData
The Weather Company
Weather Models
127K Global Stations
40M+ Mobile Phones
50K Flights a Day
Air Quality and Pollen
Global Lightening
Traffic / Incident Data
Historical
Current
Predictive
Global
Ultra-local
Weather
Atmosphere
Sources Types
National Weather
Service
The Weather Company
Condition
Updates
1 per hour
Forecast
Updates
Every 6 hours
Data and Analytic Processing
60 per second
Every 15 minutes
40M+ Mobile Phones
Handles 26 billion requests a day
3 Billion forecast reference points
Generates 4GB of new data each second
Data Ingest and Distribution
14. Italy coverage : Personal Weather Station Network
• Over 4500 Personal Weather Stations in Italy today
• 5 Minute Reporting Frequency
• Feeds into our 500 m^2 resolution Forecast on Demand Engine, especially for next 6 hours
• Historical Data back to 2005 - More sensors online as time progresses
Historic Data Points from Public Sources
Public Sources vs. Weather channel sensors
Weather.com sensors for collecting weather information
Public sources sensors for collecting weather information
Public sources sensors for collecting weather information
METAR Sensors + NWS NWS + WSI + PWS Sensors
15. Integrated Weather and Crop Modeling
Irrigation scheduling. Minimizes water wastage. (2-
10 day) forecasts.
Farm Operations. Proactive management of disease /pests. Reduce wastage
(2-10 day) forecasts.
Continuously run
weather forecast
and crop /
disease model
Generate advisories
on likelihood of rain,
Highlight conditions
for disease outbreak etc.
Advisories on fertilizer
and pesticide application.
Farmer plans activities such
as fertilizer application,
pesticide application,
harvesting,
sun drying grain.
Continuously run
weather, hydrology
models and sensors
to determine water
requirement at
paddy fields.
Use forecasts to
supply water or
drain excess.
16. Farmers edge partners with the weather company
IBM Weather Company and Farmers Edge: integration of hyper-local
forecasts from Weather’s Forecasts on Demand weather forecasting
engine into its field-centric approach to predictive modeling
IBM Weather Company Farmers Edge will provide real-time data from the
field, highly precise, predictive models to make decision on: critical crop
stages, the timing of field operations, pest and disease pressure,
equipment deployment, soil needs, and nutrient requirements