Weitere ähnliche Inhalte
Ähnlich wie Advanced Analytics & AI for Network Operations (20)
Mehr von James Crawshaw (13)
Kürzlich hochgeladen (20)
Advanced Analytics & AI for Network Operations
- 1. © 2019 Heavy Reading
Advanced Analytics & AI for Network
Operations
Telco AI Summit Europe
6 November 2019
- 2. © 2019 Heavy Reading
Panelists
2
Lucy Goudie
Customer Experience Manager
Telefonica
Dima Alkin
VP, Service Assurance Solutions
TEOCO
Moderator: James Crawshaw
Senior Analyst
Heavy Reading
George Iskenderian
Director, Big Data & AI
Bell Canada
- 3. © 2019 Heavy Reading
Use Cases for AI/ML in Telecom
3
92%
75%
67%
58%
58%
42%
42%
33%
25%
25%
25%
8%
17%
33%
42%
42%
58%
50%
58%
67%
50%
42%
8%
8%
8%
8%
25%
33%
Predictive maintenance
Security
Self-organizing networks
Network management
Fraud/revenue assurance
Data monetization
CEM
Dynamic pricing
New business models
Strategic marketing
Sales
Critical Somewhat important Not important at all
Source: Heavy Reading Survey – Thought Leadership Council, Nov. 2017
- 5. © 2019 Heavy Reading
Phased Adoption of AI/ML in Telecom Networking
5
- 6. © 2019 Heavy Reading
Type Description
Descriptive
Analytics
Summarizing data and presenting visually in a table or chart to monitor a
system or service.
Reactive
Analytics
Raising alarms based on events such as hardware failures. Uses traditional
rules and statistical methods such as clustering.
Predictive
Analytics
Predicting events and future states based on time series data. Can use
supervised, unsupervised or semi-supervised learning. Algorithms can be
linear (e.g., regression) or non-linear (neural networks, support vector
machines).
Prescriptive
Analytics
Similar to AI, where the goal is to identify an optimized sequence of steps or
actions to take. An example would be optimizing data routing across all
network nodes and end devices as traffic changes.
Analytics Classifications – From Passive to Proactive
6
- 7. © 2019 Heavy Reading
Barriers to Data Science/ML in Telecom
7
49.4%
47.9%
47.9%
33.5%
32.7%
32.3%
27.2%
22.2%
21.4%
21.0%
19.8%
17.1%
17.1%
16.7%
16.0%
Dirty data
Lack of data science talent
Lack of management/financial support
Lack of clear question to answer
Data unavailable or difficult to access
Results not used by decision-makers
Privacy issues
Integrating findings into decisions
Need to coordinate with IT
Explaining data science to others
Lack of domain expert input
Expectations of project impact
Multiple ad-hoc environments
Limitations of tools
Can't afford data science team Source: Kaggle
- 8. © 2019 Heavy Reading
TM Forum AI Maturity Model
8
Strategy
- Portfolio &
ideation
- Finance &
investment
- Strategic
management
Culture
- Culture
- Leadership &
culture
- Organizational
design & talent
management
Party
- Engagement
- Experience
- Trust
Operations
- Service
lifecycle
management
- Closed-loop
service
assurance
- Governance
Data
- Data
governance
- Data curation
- Data
capitalization
Technology
- AI Techologies
& capabilities
- AI
Architecture
- Automation &
applications
+54 specific digital criteria