Documente Academic
Documente Profesional
Documente Cultură
Telecommunications Industry
AN OVERVIEW
SUBMITTED TO:
MS USHA YADAV
NIFT, JODHPUR
PAGE 1
Future Applications of AI in Telecoms ...................................................................... 36
Conclusion ................................................................................................................... 37
Bibliography ................................................................................................................ 38
PAGE 2
ACKNOWLEDGEMENT
We, Aastha Arora and Neelesh Verma, BFT – IV, hereby take this opportunity to thank all
our fellow peers from National Institute of Fashion Technology, Jodhpur who provided
insight and expertise that greatly assisted the research.
We would also like to thank Ms. Usha Yadav, Faculty at NIFT, Jodhpur for sharing her
pearls of wisdom with us and her continuous support and guidance she rendered for the
successful completion of this project.
We express our gratitude towards our institute NIFT, Jodhpur for giving us the opportunity
to conduct this market survey.
In the end a special thanks to all members who are directly or indirectly associated with
the project.
PAGE 3
Executive Summary
The complexity of communications networks seems to increase inexorably with the
deployment of new services, such as software-defined wide-area networking (SD-
WAN), and new technology paradigms, such as network functions virtualization
(NFV). To meet ever-rising customer expectations, communications service
providers (CSPs) need to increase the intelligence of their network operations,
planning and optimization.
To help navigate the AI/ML topic, this report provides an overview of AI/ML,
outlines the key telecom use cases, quantifies the level of adoption in CSPs today,
and discusses the challenges of applying AI/ML to the networking domain
Introduction
The telecommunications sector is made up of companies that make communication
possible on a global scale, whether it is through the phone or the Internet, through
airwaves or cables, through wires or wirelessly. These companies created the
infrastructure that allows data in words, voice, audio or video to be sent anywhere
in the world. The largest companies in the sector are wireless operators, satellite
companies, cable companies, and internet service providers. (Beers, 6th February
2019)
Being at the epicenter of innovation, growth and a disruption for all industries,
telecom sector has made an irreplaceable place for itself. Mobile devices and
related broad band connectivity continue to be more and more embedded in the
fabric of society today and they are key in driving the momentum around some key
trends such as video streaming, Internet of Things (IoT), and mobile
payments.
PAGE 4
Artificial Intelligence and its Expansion
Artificial Intelligence is basically a way of making a computer, a computer-
controlled robot, or a software think intelligently, in the similar manner the
intelligent humans think.
AI is accomplished by studying how human brain thinks, and how humans learn,
decide, and work while trying to solve a problem, and then using the outcomes of
this study as a basis of developing intelligent software and systems. (Point)
PAGE 5
Figure 1s
As the market is seeing a rapid growth in Europe, North America and Latin America,
the telecom services spending in Asia-Pacific region was projected to grow by
around 2.06% in 2016 compared to 2015.
There is an increasing realization that enhancing the telecom industry with AI could
lead to various new revenue streams.
PAGE 6
Figure 2: Global revenue from telecommunications services, 2005 to 2019 [in billion Euros]
AI has vast applications in this sector; the most significant potential are the
classification of traffic, anomaly detection and network optimization and
utilization. Apart from this, it can also help in:
− Self-learning
− Self-healing
PAGE 7
IDC predicts that 31.5% of the telecommunication organizations are mainly focusing
on utilizing current infrastructure and 63.5% are committing investments on AI
systems.
While these continue to be global trends, India should equally see an increase
around AI; primarily driven by enterprise needs to drive viable efficiencies and
consumer demand for contextualization.
On the subscriber side, AI and Machine Learning will help telecom operators in
subscriber profiling and analyzing offer conversion rates, content usage
trends and network activity. This will help them push offers that are tailored as
per subscriber needs at the right time, believe analysts from their counterpoint.
Using AI and data analytics, operators will be able to identify and push various
services to the customers at the right time, for e.g. – in case of post-paid customers,
operators must encourage high speed data services and offer tailored data packs
when subscriber is running low on data. The timing of offering tailored packages
based subscriber intelligence is very important.
PAGE 8
Need for Artificial Intelligence in Telecom Industry
Figure 3
Telecom becomes the default highway for anything that is to do with digital and
adds a lot of opportunities in the telecom domain.
PAGE 9
Basic aim of Artificial Intelligence in Telecom Industry:
− network coverage,
Typically, networks through nodes observe something and then the controller,
generally a human being, takes a desired action. With AI, the network can decide
on its own and also take the next course of action through various hardware /
software solutions, essentially IoT solutions. (Kawoosa).
Machine Learning, the network will keep on adding intelligence, so it will grow in
capabilities like humans as they acquire more skills and knowledge.
PAGE 10
Existing business processes such as network operations (both planning and
optimization) have been performed manually resulting in delays and errors, which
negatively impact on customers' experience. To resolve these challenges, CSP
business processes1 can be automated using AI capabilities such as machine
learning, deep learning, and natural language processing. The need for AI to drive
automated CSP operations will continue to grow as the CSP network moves from
being physical to being virtual.
Software defined networking (SDN) and network functions virtualization (NFV) will
be dependent on automated processes to deliver service agility and cost efficiency.
Capabilities such as self-diagnostics and self-optimization can only be achieved
using intelligent insights obtained from the analysis of quality data sets.
AI-enabled networks can think beyond their correlative programming and suggest
outcome-based scenarios (‘what would you like to happen’). In the future, AI will be
able to differentiate between correlative and causal, and proactively pursue their
own choice of outcomes beyond the scope of human programming, and before any
problems are noticed by subscribers (‘I can take care of myself’).
1
communicating sequential processes is a formal language for describing patterns of
interaction in concurrent systems.
PAGE 11
Integration of AI in the Telecom Industry
There will be requirement of all software/hardware tools to add intelligence. These
will help in building sensory system to the networks and in a decentralized
architecture which is important for such a solution. Also, through SDx (Software
Defined Anything), the networks will have the agility to respond to the situations
without requiring phenomenal changes in the system components. (Kawoosa)
At a granular level, this could further help them in optimizing the profit margin
arising from enhanced network operations and reconfiguring the network to
restore or mitigate services in the event of any cyber security attack.
2
SDN: Software Defined Network Technology; is an approach to cloud computing that facilitates
network management and enables programmatically efficient network configuration in order to
improve network performance and monitoring.
NFV: Network Functions Virtualization; is a network architecture concept that uses the technologies
of IT virtualization to virtualize entire classes of network node functions into building blocks that
may connect, or chain together, to create communication services.
PAGE 12
learning (ML), and artificial intelligence (AI) are inevitable when it comes to the
industry’s future.
By using data science, machine learning, and artificial intelligence strategies,
telecommunication companies can improve four areas of their services.
The importance of data science, ML, and AI to the telecom industry will likely
present itself in these four areas in particular, which this paper will take a look at
individually:
1. Troubleshooting:
One of the major challenges for telecom providers is being able to guarantee quality
service to subscribers. Analyzing call detail records (CDR) generated by subscribers
at any given moment of the day is key to troubleshooting. However, CDRs are
challenging to work with because the volume of data gets massive and unwieldy
quickly.
For example, the largest telecommunication companies can collect six billion CDRs
per day.
With data science, machine learning (ML), and artificial intelligence (AI),
companies can instantaneously parse through millions of CDRs in real-time, identify
patterns, create scalable data visualizations, and predict future problems.
2. Fraud Detection:
Verizon estimated in 2014 that fraud costs the telecom industry upwards of $4 billion
a year. However, the faster that telecom companies analyze large amounts of data,
the better off they are in identifying suspicious call patterns that correlate with
fraudulent activity.
Cutting-edge ML and AI strategies like advanced anomaly detection make it much
easier for telecommunication companies to identify “true party” fraud quickly.
3. Marketing:
The high churn rate in telecommunications, estimated at between 20-40% annually,
is the greatest challenge for telecom companies. Telecommunication companies can
PAGE 13
use data to build better profiles of customers, figure out how to best win their loyalty
(in the most scalable and automated way), and adequately allocate a marketing
budget. With improved data architecture, they are able to harvest and store a greater
diversity of data that provide insights into each customer such as demographics,
location, devices used, the frequency of purchases, and usage patterns. By
combining data from other sources like social media, they can have a stronger
understanding of their customers.
Using machine learning gives a more accurate picture of which channels are most
responsible for customer conversions for better ad buying as well.
5. Customer Experience:
Telecommunication companies can enhance their services by analyzing the millions
of customer complaints they get every year to figure out which types of
improvements will have the greatest impact on customer satisfaction and thereby
increase customer retention. They can also leverage data at a larger and more
automated scale to gain insights into the performance of their technicians.
The more that telecommunication companies can analyze data on customer calls,
the more they can begin to recognize which types of problems are most likely to
lead to unwarranted “truck rolls” and put in place measures to prevent those calls.
Given the number of calls and the depth of analysis required, this necessarily
dictates a machine learning approach – more specifically, a deep learning approach.
PAGE 14
Because analyzing the calls themselves means dealing with lots of unstructured
data, it’s the perfect place to expand into ML and deep learning for big gains.
− Predictive maintenance
− Fraud mitigation
− Cybersecurity
− CEM
− Alert/alarm suppression
PAGE 15
− Trouble ticket action recommendations
AI/ML could use clustering to find correlations between alarms that had previously
been undetected or use classification to train the system to prioritize alarms.
Traditional rule based alarm correlation suffers from a heavy burden of rule
maintenance. With ML we could instead train a system to devise its own rules based
on a given set of data inputs (e.g., network telemetry).
PAGE 16
− Other general AI applications (RAN, core or end-to-end network):
energy efficiency according to dynamic traffic pattern, etc.; end-to-end
service orchestration and assurance (e.g., customized SLA); end-to-end
service optimization, prioritization.
PREDICTIVE MAINTENANCE
A subcategory within network operations rather than a separate field. In a 2017
survey of CSPs by Heavy Reading (see Figure), it was found that predictive
maintenance was the top use case for AI/ML in telecom, ahead of security, network
management, and fraud/revenue assurance.
FRAUD MITIGATION
Fraud detection and prevention was the fifth most popular use case in the survey
results shown above. According to the Communications Fraud Control Association,
fraud costs the global telecom industry $38 billion annually, of which roaming fraud
accounts for $10.8 billion.
PAGE 17
− Anodot uses AI to identify revenue leakage and surface discrepancies
between expected results and how events are actually billed.
CYBER SECURITY
Cybersecurity Security was the second most popular use case. Heavy Reading's
Telecom Security Market Tracker has found that there is guarded optimism over AI
for the automation of CSP security. (CRAWSHAW)
Traditional security technologies rely on rules and signatures to find threats but this
information can soon become out of date. The tactics of adversaries are evolving
rapidly, and the number of advanced and unknown threats targeting CSP networks
continues to increase. AI/ML algorithms could be trained to adapt to the changing
threat landscape, making independent decisions about whether an anomaly is
malicious or providing context to assist human experts.
One recent hot area of activity is in baselining of the behaviour of devices connected
to the Internet of Things (IoT). Here many established vendors and AI start-ups are
developing solutions that will help CSPs to manage IoT devices and services more
securely, making use of automatic profiling of those devices.
3
CSP: CSPs represent communications service provider; is a service provider that transports
information electronically
PAGE 18
CUSTOMER SERVICE & MARKETING VIRTUAL DIGITAL
ASSISTANTS
One of the key applications of AI/ML in the telecom sector to date has been the use
of chatbots to augment or replace human call center agents.
For example, Telstra's Kieran O'Meara, Director Technology Design & Delivery,
estimates that 30 percent of inbound calls to a contact center could be resolved
by AI chatbots. There is still a role for human agents at Telstra (it has 11,000 today),
but with AI assistance O'Meara estimates that they can be 20-35 percent more
productive. Telstra has around 300 agents managing chatbots on its websites but
doesn't expect this number to grow. Instead, it plans to increase the number of
agents dealing with customer enquiries directly via messaging apps such as
WhatsApp.
− Customer voice and text sentiment analysis – Telstra is looking at using text
sentiment analysis to enhance the performance of its messaging and chat
agents.
4
CRM: Customer-relationship management is an approach to manage a company's interaction with
current and potential customers. It uses data analysis about customers' history with a company to
improve business relationships with customers, specifically focusing on customer retention and
ultimately driving sales growth.
PAGE 19
CUSTOMER EXPERIENCE MANAGEMENT
Heavy Reading sees Customer Experience Management as the process of managing
"all customer touchpoints" to ensure a positive relationship with the brand.
(CRAWSHAW)
As digital touchpoints continue to grow, analytics and AI are essential tools for CSPs
to understand the health of the network, the customer journey (customer care,
billing, etc.), and real-time service quality. As such the CEM category intersects
customer service, marketing, CRM and the service assurance side of network
operations and management.
PAGE 20
Real World Examples of CSPs using AI
Although there is great interest in the topic of AI, there are few public examples of
how CSPs are using it in their operations today. Below we reference some real-world
examples in customer care, networking, and security. Figure 5 is not an exhaustive
list of all uses cases of these companies, or indeed of all CSPs using AI.
Figure 5
AT&T
According to AT&T's Inside Connections Blog, the company has been building AI
and ML systems for decades, using algorithms to automate operations such as
common call center procedures, technician dispatching, and to analyze and
correct network outages. AT&T says it is now using AI to help make its networks
more secure, self-healing and self-resilient as it prepares for the rapid growth of
video traffic, particularly on mobile networks.
PAGE 21
AT&T Labs Research has described how it is using ML to create a "virtual world"
that describes its 5G infrastructure and environment – poles, buildings, building
materials, foliage – to help determine where cell sites can be placed without
requiring a site visit and identify faults in towers. AT&T also sees AI as key to
enabling organizations to take advantage of the IoT.
AT&T has published a book called Artificial Intelligence for Autonomous Networks,
which explores the potential to transform network operations, cyber security,
enterprise services, 5G and IoT, infrastructure monitoring, traffic optimization,
customer experience and care.
The book's editor published a blog late last year (Pay No Attention to the Man
Behind the Curtain: A Reality Check for Artificial Intelligence) warning that "many
of the things we think of as AI today in fact require people to manually input and
structure massive amounts of data. The end user might be interacting with software,
but behind the scenes, there are a lot of people sweating to make that software look
smarter than it really is. For example, even commercial AI solutions that are used
today, such as virtual assistants and call routing systems, are developed on an
enormous amount of data that is labeled manually. Extensive human effort goes into
making one AI application a reality."
AT&T launched the Acumos project as a way of reducing some of the effort required
to build AI applications.
Acumos is essentially an AT&T lab project that has been spun out to the Linux
Foundation in order to apply the concept of crowdsourcing to software
development. The beta version of Acumos has code from AT&T and Tech Mahindra
and is available to download now.
The Acumos white paper explains how the development and deployment of AI
applications is currently highly time-consuming and requires expensive, specialist
talent. Acumos will provide a common framework that reduces the need for
ML "rocket scientists" and accelerates development, thereby lowering the
barriers to AI for CSPs and companies in other industries too.
PAGE 22
COLT
Colt has created a new "AI-driven networking" project called Sentio with the aim
of developing fully automated service management capabilities. The Sentio project
started last year and Colt is currently implementing a proof of concept (PoC).
− Fault prediction
− Capacity management
− Network scaling
PAGE 23
Deutsche Telekom
Instead of buying "off-the-shelf" AI systems and robots, which can be expensive,
Deutsche Telekom is developing its own AI solutions – via its own developer teams
and partners. One area that is ripe for automation with AI is resolving queries from
enterprise customers, which Deutsche Telekom notes can sometimes require 1,000
manual actions in various software systems (entering bookings and process
commands, initiating orders, etc.).
Tinka is a chatbot for the Austrian market that has learned more than 1,500 answers
so far. Tinka is able to handle about 80 percent of all questions put to her. When
she can't answer a question, she forwards it to a human. Tinka can support
customers in setting up LTE-based home WiFi networks and inserting SIM cards
into phones.
PAGE 24
Vanda is a natural language processing chatbot used for enterprise self-service in
Deutsche Telekom's eastern European markets (e.g., Hungary) via Facebook, Viber
and other messaging and voice platforms.
PAGE 25
KDDI
In 2016 the R&D Labs of Japanese operator KDDI announced an AI-based
monitoring system that was used to predict anomalies caused by NFV hardware and
software. The AI-based monitor learns what are normal and abnormal the
conditions, and the software-defined networking (SDN)/NFV orchestrator provides
an appropriate recovery plan so that affected services are quickly preserved. KDDI
described the PoC as a world first and an important step toward the network
virtualization era supporting 5G.
Figure 6
PAGE 26
SK Telecom
In October 2017, SK Telecom announced the expansion of its T Advanced Next
Generation Operational Supporting System (TANGO) from the fixed to the mobile
side of its business. TANGO is an AI-assisted network operation system with big
data analytics and ML capabilities that detects issues on the network, troubleshoots
problems, and the optimizes performance. SK Telecom has also signed a strategic
partnership with Indian operator Bharti Airtel for the use of TANGO.
Telefónica
At Mobile World Congress, Telefónica presented its "fourth platform," originally
announced a year earlier. The first three platforms (physical assets, IT systems and
products/services) are table stakes for CSPs; Telefónica's fourth platform concept is
akin to the "digital service provider" model other CSPs are striving toward.
(Company)
The "fourth platform" uses AI to analyze data from the underlying platforms to
better serve customers. Specifically, Telefónica has launched its voice-activated
"cognitive" assistant, Aura, in six markets – Argentina, Brazil, Chile, Germany,
Spain and the U.K. Telefónica says that Aura will learn from its interactions with
individual customers and ultimately be able to provide tailored recommendations
and support based on a user's preferences.
PAGE 27
Telefónica is not only applying big data analytics and ML to its customer-facing
activities but also to improve the operational efficiency of the business. In 2007
Telefónica first started applying algorithms and ML to troubleshooting in its
network operations centers (NOCs). The team of data scientists from Telefónica
R&D that developed various tools for NOCs have published several academic papers
and filed multiple patents related to their work. Leveraging this small group of data
scientists is a team of hundreds that apply business intelligence and data
visualization tools to operational and commercial use cases across Telefónica.
The first part of the process is creating the data repositories themselves. Telefónica
has a separate data lake in each operating business (Spain, Germany, U.K.,
Brazil, etc.) and a centralized, global big data platform for analysis across the group.
In total Telefónica collects data from more than 170 sources of information,
including contact center calls, field technician reports, bills, energy usage, OSS,
network telemetry and so on. Over time, more and more data sources are being
added.
Once data has been collected and anonymized, it must be normalized using a
standard data model and checked for quality. Poor data from inventory systems has
required Telefónica to replace some systems (e.g., transport and access inventory)
and to change some operational processes to improve the quality of the data that is
entered into these systems. Even if a perfect network inventory is created, it will
soon diverge from reality if field technicians fail to report changes such as a change
in port on a device to resolve a trouble ticket.
Once a reliable data set is available, the focus shifts to analyzing use cases. These
normally come from business units looking for solutions to real-world problems. So
far, Telefónica's data analytics team has worked on solutions for around 300 use
cases, the benefits of which it tracks on an ongoing basis. Most of the use cases are
operations related (e.g., infrastructure management, customer experience,
customer service delivery, internal plant management, etc.). The next largest
category is technology-related (video platform, radio planning, etc.).
PAGE 28
Figure 7
Another use case where the analytics team has employed AI is to create a real-time
index of customer satisfaction. Just measuring traditional key performance
indicators (KPIs), such as dropped calls and throughput, does not always correlate
well with customer experience (as determined by survey data), especially for
complex services such as VoLTE and IPTV. As such, Telefónica has turned to some
sophisticated ML algorithms that use network KPIs collected every 15 minutes
to predict the customer's satisfaction level with an accuracy of around 60 percent
(target 75 percent by year-end).
Vodafone
Vodafone introduced the first live chatbot in the U.K. telecom market, TOBi, in
2017. Vodafone claims the AI-enabled bot, based on IBM's Watson technology,
provides relevant support to resolve more than 70 percent of customer queries.
PAGE 29
Vodafone is also trialing a voice biometrics system and a voice assistant that will
be compatible with Amazon's Alexa software and Echo speaker.
Figure 8
PAGE 30
Figure 9
PAGE 31
Figure 10
Bharti Airtel in November 2017 partnered with software firm Amdocs to deploy
machine learning and AI-based technologies across Airtel’s multiple lines of
business. Airtel said it would “leverage Amdocs’ innovation centers, delivery
expertise and its ecosystem of startups to help pre-empt and self-heal operational
issues, introduce smartbots into digital channels, and quickly launch and activate
new services, thereby enabling a seamless customer experience.”
Last year, Airtel rolled out its digital innovation program Project Next aimed at
transforming customer experience across all its services and touch points. It also
plans to launch several digital innovations to simplify and enhance interactivity of
the Airtel customer experience. Airtel said Project Next complements Airtel’s
massive investments in building a future-ready network under Project Leap for
which the telecom major has committed an investment of 600 billion rupees in
the next 3 years.
PAGE 32
Reliance Jio – ‘Data Is the New Oil’
Figure 11
Reliance Jio is on a mission to include the 500 million feature phone users of India
(out of 780 million) who have been left out of the digital revolution. “We shall
achieve this in three ways – connectivity, data affordability and device affordability,”
Mukesh Ambani, chairman and managing director of Reliance Industries, said in a
letter to shareholders.
“Data is the new oil,” Ambani said in the first edition of the India Mobile Congress
held in New Delhi in September 2017. India needs to instil confidence in new
technologies such as artificial intelligence, robotics, IoT and big data analytics,
which will be the enablers of a new wave of wealth and employment creation, he
added.
Jio has introduced voice commands on its MyJio application to ease the customer
service process. Bank of America Merrill Lynch said in a recent report that voice AI
is a focus area for the Mukesh Ambani-led telco and the AI-based voice command
feature is working well for the telco.
PAGE 33
Brands are constantly seeking new ways to interact with consumers. For them, here's
something unlike anything seen before - it's called the 'Video Call Bot' and it's been
created by the mobile network Jio.
It was used recently for the promotion of a new film, 102 Not Out, featuring Amitabh
Bachchan and Rishi Kapoor. Bachchan's bot was later followed by Ranbir Kapoor for
the movie Sanju and Baba Ramdev for day-to-day yoga tips.
Jio claims that its Video Call Bot is an "innovation, that's a first in the world". No
tech company has ever created an AI-based Video Call Bot, the brand states.
JioInteract is a patented AI-based brand engagement platform and as for the Video
Call Bot, it mimics a human by creating vocalised responses. In the recent cases
of interaction with the Video Call Bot, the conversations with Bachchan were both
in Hindi and in English.
The platform can be accessed from the MyJio application available on both the
Google Play Store and the Apple App Store.
PAGE 34
CHALLENGES OF APPLYING AI/ML TO NETWORKING
Generic AI applications such as marketing and chatbots apply to many industries,
not just telecom. When it comes to applying AI/ML to networking, the industry is
at a more nascent stage, partly because these problems are specific to
communications, and partly because there is a lack of public data sets on which
academics have been able to experiment.
In the paper Big Data Analytics, Machine Learning and Artificial Intelligence in
Next-Generation Wireless Networks, the authors highlight the challenges of
adopting big data analytics and AI in the next-generation communication system:
− The prime concerns for the MNOs emerge from the extent of effort, skills,
and workforce needed to manage and operate a big data platform.
− However, the most important and difficult challenge is more likely to stem
from the loss of direct control that the MNOs still have over the wireless
network. The loss of direct control is incurred from the combination of
automation and real-time operations within the big data analytics
framework.
Below we detail several barriers to applying data science and ML in telecom (see
Figure). Some of these are generic to any industry, but some are more specific to
telecom.
PAGE 35
Figure 13: Challenges
As telecoms are already using AI to solve issues related to customer care, network
coverage, billing and service/product offering.
PAGE 36
Conclusion
AI and ML are not new topics. Even in the context of telecom, as far back as 1993
researchers were exploring AI techniques that they thought would be essential to
the transformation of the telecom network. The key factors that have led to an
acceleration in progress in AI/ML in recent years include breakthroughs in neural
network theory, the availability of massive data sets for academics to experiment
with, and the rise of public cloud (AWS, Azure, etc.) making computing capacity
readily available and cheap.
PAGE 37
Bibliography
Adams, B. (2016). Artificial Intelligence for Telecommunications Applications.
PAGE 38