success story

Artificial Intelligence in the Financial Sector

Industry
Solution
Deep Learning
Focus for
Area
Cryptocurrency Trading

Using Sentiment Analysis to

Gauge Cryptocurrency Value
Deep learning provides a way to analyze sentiments about cryptocurrencies by
scanning and evaluating comments across the web, including news headlines,
Twitter posts, and Reddit posts.

“I have learned that there is Challenge

correlation between sentiment Determining the valuation of cryptocurrencies is difficult because the worth does
not tightly correspond to factors such as cash flow or available assets, as it does in
and cryptocurrency prices. the conventional stock market. With dozens of currencies available in the market, a
This is something that may be system is needed to methodically evaluate their worth.

helpful to other developers: Solution

to explore new industries Emerging neural network models, including recursive neural tensor networks
(RNTN), provide a promising mechanism for determining the feelings about any
and see how current AI given currency by scanning and parsing comments in the social media. Favorable
technologies can be applied sentiments about a currency can be shown to correspond with an uptick in the
currency’s value across digital coin exchanges.
to create a solution or insight
within a new area.” Background and Project History
Teju Tadi, an Intel AI Ambassador with a strong interest in blockchain and
—Teju Tadi, Artificial Intelligence Ambassador cryptocurrencies, launched a project in May 2017, applying deep learning
Program, Intel Corporation techniques to investigate the correlation between trader sentiments for
cryptocurrencies and their market value. Sentiment analysis as a way of helping
gauge trading trends has already gained adherents in the traditional stock market.
Because the valuation data available for cryptocurrencies is more nebulous, Teju
is refining techniques to combine trader sentiments with other factors to create
better ways to anticipate trends.
“Many firms in the equities space have been employing techniques,” he said, “which
make key investment decisions based on social media data and news headlines.
There are algorithms that make decisions to instantaneously buy a certain equity
as soon as positive news is released—faster than any one person could. The same
methodology, I thought, could be applied to the cryptocurrency space, because
a lot of these currencies are sentiment driven. Social media and news inevitably
affect the prices of various currencies greatly.”
“I first got interested in blockchain technology and cryptocurrencies back
when I was high school,” Teju said. “At that time there were fewer than 100
cryptocurrencies and the industry was still in its infancy. The market evolved once
the Ethereum blockchain protocol launched. Ethereum and other platform tokens
enabled developers to deploy software for blockchain protocols much faster. As
of today, there are over 1600 cryptocurrencies and blockchain protocols and many

develop
projects still launching.”

the future of ai for All

Teju’s interest in machine learning and deep learning grew generate increasing value in the long term. The tamper-
as he began working on projects outside of school. A project resistant nature of the blockchain’s historical entries—
using machine-learning strategies, ProvidR*, won first place logically linked together in a continuous chain—provide
in the Google Community Leaders Program (CLP) Case a mechanism to securely authorize and log transfers of
Competition. ProvidR offered a scalable solution to food cryptocurrencies from one party to another.
insecurity, aggregating requests by people who visited a
Even as the potential of blockchain as a secure digital
particular food bank. He then began a project working with
alternative to traditional financial processes is being
Intel, Face It*, which employed deep learning techniques to
realized, the underlying blockchain architecture—enabling
map an individual’s facial structure and then recommend a
distributed database capabilities—presents opportunities in
hairstyle suited to their appearance. This early work led to
other applications. Enterprises are exploring ways in which
his current endeavor, Deep Learning for Cryptocurrency
blockchain could be used as a part of Internet of Things
Trading, that capitalizes on the expertise he has gained in
(IoT) solutions to manage and gain insights into supply chain
finance, blockchain architecture, cryptocurrency, and AI.
activities and global operations. Blockchain architecture
could also be used in healthcare information systems to
distribute information from medical devices and consolidate
and distribute an individual’s healthcare records

“Because blockchain is based on a distributed,

peer-to-peer topology where data can be stored
globally on thousands of servers—and anyone
on the network can see everyone else’s entries
in real-time—it’s virtually impossible for one
entity to gain control of or game the network.”1
—Lucas Mearian, Senior Reporter, Computerworld

Commercial Opportunities
In an article for the Intel® Software Developer Zone, Teju
stated, “The long-term vision of this project is to be able to
develop an AI cryptocurrency trading bot that can not only
consider trader sentiment to make trading decisions but also
take advantage of other opportunities such as arbitrage,
which is the purchase and sale of an asset to profit from a
difference in the price.”
Teju took the insights gained from his research as he helped
Figure 1. Teju Tadi presenting his project at the Intel® AI establish a business, Mycointrac, focused on providing
DevCon 2018 (Intel® AIDC) in San Francisco. cryptocurrency market intelligence. “Once the product
is fully developed,” he said, “I plan to utilize the data
“Working with Intel’s AI Student Ambassador program, Teju provided by it as one of the factors to make key investment
said, “has opened so many opportunities for me. It’s allowed decisions for my new cryptocurrency hedge fund, Sentience
me to connect with peers working in AI space across the Investments L.P., which has been operational since January
world. It gives me access to Intel engineers who are more first. The plan is to develop trading strategies based on a
than willing to help me on my projects. Intel also provides number of high-frequency, machine-learning techniques, as
me with access to hardware which helps me test and deploy well as deep learning and sentiment analysis.”
any applications that I am working on. Most of all it gives
me a great environment, which fosters and enables me to Each individual exchange, Teju explained, has its own
pursue my AI-related interests and projects.” supply and demand and its own set of buyers and sellers.
The market as it stands is very inefficient. A cryptocurrency
Watch a video showing Teju talking about his project.
in a certain exchange might be trading at USD 3 and then,
in another market, it will be trading at USD 5, for the same
The Evolution of Cryptocurrencies and Blockchain
currency. Traders use these differential valuations to
Cryptocurrencies serve as a medium for exchanging digital advantage by executing arbitrage—buying on the exchange
assets; transactions within this medium are recorded using where it is trading for USD 3 and then selling it at the
blockchain techniques, which operate as an encrypted, exchange where it is trading for USD 5 for a riskless profit
electronic ledger providing a permanent history of all of USD 2.
activities. Because cryptocurrencies are issued based on
a finite supply, investors anticipate that rising demand will

2
“In Mycointrac, if you click on one of the coins on the market and scroll,” said Teju, “you’ll see the gap price versus the index.
That, basically, is arbitrage. That is showing you how much it is trading up or down at a certain exchange. Sometimes
it is like a few percent. Sometimes it could be 50 percent or 100 percent. You will easily see 5 to 10 percent differences on
most coins.” Figure 2 shows live sentiment tracking displayed in Mycointrac. Working with a team, Teju is taking insights from
this project to launch a new site, Mycoinrisk, focusing on fraud prevention and risk mitigation in the cryptocurrency space.

Figure 2. Live sentiment mapping on MyCointrac gauges cryptocurrency chatter on the web.

“As machine learning and artificial intelligence surrounding the cryptocurrency. Based on these reasons,
I believe that sentiment analysis of news headlines, Reddit
(AI), applications continue to increase and posts, and Twitter posts should be the best indicator of the
impact accounting and finance responsibilities, direction of cryptocurrency price movements.”

the human professionals have an opportunity Recurrent neural networks (RNNs) have been a prominent
technique for sentiment analysis, Teju noted. RNNs parse
as well. Not only will they be more productive a string of text and tokenize the words, determining the
and proficient, but they will be able to handle frequency of words used and creating what is called a
bag-of-words model, often used in document classification
more clients and deliver more value because with word frequency being used to train a classifier. The
subjectivity of each word is searched from a lexicon in which
they can determine actionable insight rather emotional values were prerecorded by researchers. From
than just crunch numbers. Machines will be this data, the overall sentiment is gauged.
able to propel innovation in the industry.”2 “RNNs work well for longer texts,” Teju said, “but are
—Bernard Marr, Author and Keynote Speaker ineffective at analyzing sentiment in shorter texts, such as
on Business and Technology news headlines, Reddit posts, and Twitter posts. RNNs fail to
consider all the semantics of linguistics by failing to consider
As is the case with all AI projects, models are continually compositionality—the order of words in a string. Because of
refined over time, with iterative training to strengthen the this, RNNs are ineffective at identifying change in sentiment
results and improve precision of the output. Teju explored and understanding the scope of negation.”
several neural network models before deciding on an RNTN
as the most effective way to perform natural language Recursive Neural Tensor Network
processing of social media feeds and news items. After considering the alternatives, Teju decided that an
“Many of these cryptocurrency price movements,” Teju RNTN would be the best option for his project because
said, “could be determined by herd instinct. Herd Instinct, of its capability of being able to assess the semantic
according to behavioral finance, is a mentality characterized compositionality of text. For shorter pieces of text, such as a
by lack of individual decision making, causing people to tweet, the compositionality is vital to being able to accurately
think and act in the same way as the majority of those determine sentiment from a sparse set of information.
around them. The price movements tend to be based on
market sentiment and the opinions of the communities

3
“RNTNs,” Teju said, “are great at considering syntactical Data that is ingested by the sentiment analyzer is parsed
order. RNTNs are made up of multiple parts including the into a binary tree. Specific vector representations are
parent group known as the root, the child groups known formed of all the words and are represented as leaves. From
as the leaves, and the scores. Leaf groups receive input the bottom up, these vectors become the parameters to
and the root group uses a classifier to determine the class optimize and serve as feature inputs to a softmax classifer*.
and score.” Vectors are classified into five classes and assigned a score.
“The next step,” Teju said, “is where recursion occurs.
Recursive neural tensor networks (RNTNs) When similarities are encoded between two words, the two
are neural nets useful for natural-language vectors move across to the next root. A score and class are
outputted. A score represents the positivity or negativity
processing. They have a tree structure with of a parse while the class encodes the structure in current
a neural net at each node. You can use parses. The first leaf group receives the parse and then the
second leaf receives the next word. The score of the parse
recursive neural tensor networks for boundary with all three words are outputted and it moves on to the
segmentation, to determine which word groups next root group.”

are positive and which are negative. The same “The recursion process continues until all inputs are used
up, with every single word included. In practical applications
applies to sentences as a whole. RNTN’s end up being more complex than this. Rather than
using the immediate next word in a sentence for the next leaf
Word vectors are used as features and serve group, an RNTN would try all the next words and eventually
checks vectors that represent entire sub-parses. Performing
as the basis of sequential classification. They this at every step of the recursive process, the RNTN can
are then grouped into subphrases, and the analyze every possible score of the syntactic parse.”

subphrases are combined into a sentence that Figure 3 shows an example of how a sentence is parsed and
analyzed using an RNTN approach.
can be classified by sentiment and other metrics.
More details of Teju’s techniques in this project can be found
—Excerpt from Recursive Neural Tensor Network, Deeplearning4J on Intel® Developer Zone in the paper Deep Learning for
Cryptocurrency Trading.

20²⁄ ₃

13 20¹⁄ ₃

Still
13 20²⁄ ₃

,
13 20

13 13 20¹⁄ ₃ 13

this flick .
13 21

is
19²⁄ ₃ 19¹⁄ ₃

19²⁄ ₃ 13 13 18¹⁄ ₃

and host
20¹⁄ ₃ 13 13 19

fun , to
14²⁄ ₃ 23¹⁄ ₃

some
24 14¹⁄ ₃

sequences
17 23

truly excellent
Figure 3. Example of scoring from the Stanford Treebank.

4
SUCCESS STORY | Using Sentiment Analysis to Gauge Cryptocurrency Value

Enabling Technologies as there are many libraries, tutorials, technical articles, and a
plethora of digital content that you could learn from.”
Teju gave a nod to the benefits of using of Intel® technologies
in his projects. “My solution utilizes Intel® AI Devcloud, Intel® Other valuable resources for those getting started with
Distribution for Python*, and Intel® Optimization for Caffe*,” AI include:
he said.
Video: What is Intel Optimized Caffe
Intel AI DevCloud served as the development platform
Article: Manage Deep Learning Networks with Caffe
during the early stages of the project. “At the start of the
Optimized for Intel Architecture
project, for the sandbox version, I used Intel AI DevCloud
to run the recurrent neural networks and experiment with Article: Get Started with Intel Distribution for Python
Twitter data to see how the models were working. For my
initial project with Intel and for my current project with Intel, AI is Providing New Opportunities in the
it is completely using Intel AI DevCloud and its supporting
technologies.”
Financial Sector
Through the design and development of specialized chips,
Intel AI DevCloud, a server cluster featuring Intel® Xeon® sponsored research, educational outreach, and industry
Scalable processors, available to Intel® AI Developer Program partnerships, Intel is firmly committed to advancing the state
members free of charge, is preloaded with frameworks and of AI to solve difficult challenges in medicine, manufacturing,
tools to quickly launch machine learning and deep learning agriculture, scientific research, and other industry sectors.
projects. Pre-installed components include neon™ framework, Intel works closely with government organizations, non-
Intel® Optimization for Theano*, Intel® Optimization for government organizations, educational institutions, and
TensorFlow*, Intel® Optimization for Caffe*, and the Keras* corporations to uncover and advance solutions that address
library. Connections, once approved, take only about 10 major challenges in the sciences.
minutes to set up, by means of a Linux® terminal or graphical
user interface client, such as PuTTY*. Access through For example, consistently gaining high returns on stock
Microsoft Windows* is also supported. At this point, you’re market investments has been phenomenally difficult, even for
ready to begin training models or running Python code. For very experienced investors. Bringing AI techniques to bear on
a thorough introduction to the process, read Getting started this challenge, an international team devised algorithms using
with the Intel AI DevCloud. Deep learning can be a challenge past market data to simulate real-time investment. These
for those just getting started, but fortunately there are techniques demonstrated a 73 percent return on investment
many resources for gaining an understanding of models and compared with the 9 percent typical of real market scenarios.
initiating training. The algorithms proved particularly effective during times of
extreme market volatility, suggesting that AI can detect and
Connections with libraries, code, and examples also proved respond to patterns that human investment managers fail to
invaluable during the project development. “Intel Developer recognize. The lead author of the study was Dr. Christopher
Zone has been a great resource to learn about how I could Kraus, chair for Statistics and Econometrics at the School of
utilize Intel technologies to better build my product,” Teju Business and Economics at Germany’s Friedrich-Alexander-
said. “There are also a great number of projects out there Universität Erlangen-Nürnberg.4
built by peers in the AI industry from which I got both
motivation and valuable insight into the various ways AI
and ML were being used in a wide range of industries and
use cases.”
“I recommend checking out the Stanford Sentiment
Treebank to learn about recursive neural tensor networks.
I also recommend taking a look at the Intel Developer Zone

“AI’s role in finance may not attract as much attention in Hollywood, but it is likely to have a far
greater economic impact than consumer tech. From extending investment opportunities to the
underbanked to thwarting fraud to mitigating investment risks, AI has the potential to not only
revolutionize the industry, but also to improve the financial health of millions of people in the US
and across the world.”5
—Kevin Dinino, President of KCD PR

5
The Intel AI technologies used in this implementation included:
“At Intel, we’re encouraged by the
Intel Xeon Scalable processors: Tackle AI impact that AI is having, driven by
challenges with a compute architecture
optimized for a broad range of AI workloads, its rich community of developers.
including deep learning. AI is mapping the brain in real
Framework optimization: Achieve time, discovering ancient lost
faster training of deep neural cities, identifying resources for
networks on a robust scalable
infrastructure. lunar exploration, helping to protect Earth’s
oceans, and fighting fraud that costs the world
For Intel AI Developer Program members,
the Intel AI DevCloud provides a cloud
billions of dollars per year, to name just a
platform and framework for machine few projects. It is our privilege to support this
learning and deep learning training.
Powered by Intel Xeon community as it delivers world-changing AI
across verticals, use cases, and geographies.”6
Scalable processors, the Intel AI DevCloud is available for
—Naveen Rao, Vice President and General Manager, Artificial
up to 30 days of free remote access to support projects by
Intelligence Products Group, Intel
Intel AI Developer Program members.

Join today at: software.intel.com/ai/sign-up

For a complete look at our AI portfolio, visit
ai.intel.com/technology.

RESOURCES
Intel® AI Developer Program software.intel.com/ai
Inside Artificial Intelligence: Next-level computing powered by Intel AI ai.intel.com/
Deep Learning for Cryptocurrency Trading. Details the sentiment analysis techniques
for evaluating cryptocurrency trading
software.intel.com/en-us/articles/deep-learning-for-cryptocurrency-trading
Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank
nlp.stanford.edu/~socherr/EMNLP2013_RNTN.pdf
Intel® AI DevCloud. Free cloud compute for Intel AI Developer Program members
software.intel.com/ai/devcloud
Intel® Developer Zone. Provides libraries, tutorials, technical articles, and digital content for developers
software.intel.com/
Intel® Software Innovator Program. Supports innovative, independent developers
software.intel.com/intel-software-innovators/overview
Intel® Optimization for Caffe*
software.intel.com/ai/frameworks/caffe
Intel® Math Kernel Library (Intel® MKL
software.intel.com/mkl

6
SUCCESS STORY | Using Sentiment Analysis to Gauge Cryptocurrency Value

End Notes
1
Mearian, Lucas. What is blockchain? The most disruptive tech in decades. Computerworld, May 2018.
https://www.computerworld.com/article/3191077/security/what-is-blockchain-the-most-disruptive-tech-in-decades.html
2
Marr, Bernard. The Digital Transformation of Accounting and Finance – Artificial Intelligence, Robots, and Chatbots. Forbes, June 2018.
https://www.forbes.com/sites/bernardmarr/2018/06/01/the-digital-transformation-of-accounting-and-finance-artificial-intelligence-robots-and-chatbots/2/#61c9039417d8
3
Barsolai, Chris. The Complete Developer’s Guide to Intel AI Resources and Tools. Medium. October 2017.
https://medium.com/intel-student-ambassadors/the-complete-developers-guide-to-intel-ai-resources-tools-portfolio-530b01c7b4a6
4
Walters, Greg. Artificially Intelligent Investors Rack Up Massive Returns in Stock Market Study. Seeker. March 2017.
https://www.seeker.com/artificially-intelligent-investors-rack-up-massive-returns-in-stock-ma-2321650774.html
5
Dinino, Kevin. Five ways AI is disrupting financial services. FinTech Futures, April 2018. https://www.bankingtech.com/2018/04/five-ways-ai-is-disrupting-financial-services/
6
Rao, Naveen. Helping Developers Make AI Real. Intel, May 16, 2018. https://ai.intel.com/helping-developers-make-ai-real/

Notices
Intel technologies’ features and benefits depend on system configuration and may require enabled hardware, software or service activation. Performance varies depending on system
configuration. Check with your system manufacturer or retailer or learn more at intel.com.
No license (express or implied, by estoppel or otherwise) to any intellectual property rights is granted by this document.
Intel disclaims all express and implied warranties, including without limitation, the implied warranties of merchantability, fitness for a particular purpose, and non-infringement, as well as any
warranty arising from course of performance, course of dealing, or usage in trade.
This document contains information on products, services and/or processes in development. All information provided here is subject to change without notice. Contact your Intel representative
to obtain the latest forecast, schedule, specifications and roadmaps.
The products and services described may contain defects or errors known as errata which may cause deviations from published specifications. Current characterized errata are available on request.
Copies of documents which have an order number and are referenced in this document may be obtained by calling 1-800-548-4725 or by visiting www.intel.com/design/literature.htm.
Intel, the Intel logo, neon, OpenVINO, and Xeon are trademarks of Intel Corporation in the U.S. and/or other countries.
*Other names and brands may be claimed as the property of others.
© 2018 Intel Corporation 0618/KP/MESH/PDF 337409-001US