Machine learning in finance has become more prominent recently due to the availability of vast amounts of data and more affordable computing power.
Machine learning in finance is reshaping the financial services industry like never before.
Leading banks and financial services companies are deploying AI technology, including machine learning (ML), to streamline their processes, optimise portfolios, decrease risk and underwrite loans amongst other things.
Here in this article, we will explore some important ways machine learning is transforming the financial services sector and examples of real applications of machine learning in finance.
To answer this question and understand the role of machine learning in finance, we must first understand why machine learning is suitable for finance.
Why Machine Learning is Suitable in Finance?
Machine learning is about digesting large amounts of data and learning from that data in how to carry out a specific task, such as distinguishing fraudulent legal documents from authentic documents.
Machine learning in finance is the utilization a variety of techniques to intelligently handle large and complex volumes of information.
ML excels at handling large and complex volumes of data, something the finance industry has in excess of.
Due to the high volume of historical financial data generated in the industry, ML has found many useful applications in finance.
The technology has come to play an integral role in many phases of the financial ecosystem, from approving loans and carrying out credit scores, to managing assets and assessing risk.
The following are some of the current applications of machine learning in finance.
Portfolio Management – Robo-Advisors
Robo-advisors are a common application of machine learning in finance.
Robo-advisors are an online application that provides automated financial guidance and service.
They provide portfolio management services that use algorithms and statistics to automatically establish and manage the investment portfolio of a client
These digital investment platforms simplify the investment process which can be daunting for many people.
These services are also much cheaper than consulting a human financial advisor.
In addition, many of them don’t require account minimums at all or only require low account minimums.
Optimising your investment portfolio with machine learning
To set up an account with a robo-advisor, you complete a questionnaire about your financial situation and investment needs (for instance you might want to retire at 65 with $200,000 in savings or save for your child’s college tuition).
The robo-advisor then allocates your assets across a range of investment options (e.g. stocks, bonds, real estate) based on your specific goals and risk tolerance profile and uses algorithms to monitor and periodically rebalance your portfolio.
These companies are online financial advisors that use technology to help people manage their finances.
Betterment uses algorithms to suggest an appropriate asset allocation for investors.
This is based on the answers that investors give to questions like, how do you plan to use the money, and what is your time frame.
Wealthfront leverages the impersonal advantage of technology to offer their investment services.
Their software is programmed to follow and execute proven investment strategies, to automatically look for better investment opportunities, while keeping the optimal investment mix over time.
The Wealthfront software can implement a variety of strategies, including tax-loss harvesting, which lowers the tax investors pay.
SoFi is an online lending platform that offers home improvement loans, credit card consolidation, student loan refinancing, mortgage refinancing as well as investment management.
The company offers software-based financial advice via its automated investing platform as well so-called active investing for those who want to have a go at investing and trading themselves.
Nutmeg is one of the largest digital wealth manager in the UK.
The Nutmeg robo-advisor uses information about an individual’s financial goals and risk tolerance to allocate funds to a diversified portfolio.
The robo-advisor automatically reinvests any dividends on investments and automatically rebalances a portfolio as needed.
Algorithmic trading (or simply algo trade) is the use of algorithms to conduct trades autonomously.
This is another example of how companies make use of machine learning in finance.
In algorithmic trading, computers execute programmes with a predetermined set of instructions (an algorithm) for placing a trade on behalf of a trader.
These instructions usually involve parameters like timing, price, quantity or other constraints.
Algorithmic trading enables the execution of a large order by sending small increments of the order, called ‘child orders’ to the market at intervals.
So, it’s mostly hedge fund managers that make use of automated trading systems and so make use of machine learning in finance.
Machine learning is integral to the advantages of algorithmic programs.
It allows traders to automate certain processes ensuring a competitive advantage.
The system also makes it possible to operate in multiple markets, increasing trading opportunities.
In addition, the algorithms are able to learn and adapt to real-time changes, which is another competitive advantage for those institutions that adopt machine learning in finance.
The algorithmic systems involved here are a phenomenal aid to traders.
For example, algorithms are not sentimental or emotional, which are attributes that so often sabotage human aspirations when it comes to investments.
Algorithmic trading, therefore, simplifies the decision-making process by sidestepping human emotions.
This is a crucial benefit of employing machine learning in finance.
An AI driven hedge fund that makes stock trades without human intervention is the ultimate application of machine learning in finance.
Hong Kong-based Aidyia uses algorithms to conduct trades autonomously.
AI expert and founder and CEO of SingularityNET, a blockchain-based AI marketplace, Ben Goertzel, is the company’s chief scientist.
Aidyia runs a hedge fund that uses artificial intelligence to the exclusion of humans to make all stock trade decisions.
Humans built the system, but the system runs completely on its own with no human interference.
At the launch of the automated hedge fund Goertzel famously remarked: “If we all die, it would keep trading.”
The company uses a number of AI capabilities, including one inspired by genetic evolution and another one by probabilistic logic, to make predictions about the market and conduct trades on their own.
This is not a new development.
Artificial intelligence and machine learning have been part of many hedge fund strategies for many years.
But, this is the first completely autonomous hedge fund.
Tech Revolution explains the system basically works to find the ultimate smart trader population, by continually testing the performance of their digital stock traders.
The system only retains the “genes” of the best performers to create a team of unbeatable traders. And this process continues indefinitely.
READ MORE: Pioneer AI Hedge Fund – DE Shaw
Can individuals leverage machine learning algorithms to trade successfully?
Although algo trade simplifies matters for traders and fund managers, writing an electronic trading algorithm is an incredibly complicated undertaking.
A JP Morgan analyst points out that even a medium frequency electronic trading algorithm that reconsiders its options every second requires 3,600 steps per hour.
According to Sigmoidal, a Warsaw machine-learning consultancy, it will be difficult for an individual to successfully implement an ML investment strategy.
The reason for this is that one would need access to highly talented professionals with expertise in trading and data science to develop a trading algorithm.
In this respect, large investment banks and other financial institutions stand to benefit more from machine learning in finance than individuals.
Machine Learning hedge funds already significantly outperform generalized hedge funds, as well as traditional quant funds, according to a report by ValueWalk.
According to the July 2018 edition of the Hedge Fund Sentiment Survey, more than half of hedge fund managers use AI/ML to inform investment decisions;
two-thirds use AI/ML to generate trading ideas and optimize portfolios and more than a quarter use automation to execute trades.
Well over half have used AI for three or more years, and a third have used the technology for five-plus years.
High-Frequency Trading (HFT)
Algorithmic trading that happens at high speeds beyond human capability gave rise to high-frequency trading (HFT) – that is hundreds of thousands of trades per day executed by complex algorithms that analyze multiple markets to execute orders based on market conditions.
HFT is a subset of algorithmic trading and an excellent use case of machine learning in finance.
Investment banks and hedge funds leverage automated trading platforms and algorithms that are able to track multiple financial markets to execute vast orders.
Algorithms make it possible for these big players to benefit from minute price differences that might exist only for a fraction of a second.
Machines in charge of HFT is nothing new.
During 2009-2010, anywhere from 60% to 70% of U.S. trading was attributed to HFT.
This means most investment banks, pension funds, mutual funds, and hedge funds make use of HFT.
Some of the biggest players include companies like Tokyo-based Nomura Securities, Virtu Financial, Two Sigma Securities, Citadel Securities, Tower Research Capital and DRW, but there are many more operating in financial markets worldwide.
These companies are all optimizing the capabilities of machine learning in finance.
One of the criticisms against the practice of HFT it that it can cause inexplicable and sudden market movements.
The dramatic drop in the Dow Jones Industrial Average on May 6, 2010 (10% in just 20 minutes) was afterwards blamed on a massive order that triggered a sell-off and caused the crash.
Fraud is a massive problem for financial institutions and one of the foremost reasons to leveraged machine learning in finance.
Fraud losses incurred by banks and merchants on all credit, debit, and prepaid general purpose and private label payment cards issued globally amounted to £16.74 billion ($21.84 billion) in 2015, according to a Bloomberg report.
Machine learning is ideally suited to combating fraudulent financial transactions.
This is because ML systems can scan through vast data sets, detect unusual activities, (anomalies), and flag them instantly.
ML is also the perfect candidate to tackle the problem of false positives, which is something that happens regularly in finance.
False positives, also known as “false declines”, happens when merchants or financial institutions wrongly decline legitimate financial transaction requests.
This usually happens because there are reasons to suspect fraud.
False-positive card declines are a huge pain point for financial institutions which stand to lose out on customer loyalty when a company incorrectly declines customers’ cards.
In 2015 Javelin Strategy and Research reported that at least 15% of all cardholders had at least one transaction incorrectly declined in the previous year, which represented an annual revenue loss totaling nearly $118 billion.
What’s more, 39% of cardholders whose cards were declined, said they gave up their card after it was falsely declined.
For companies, this means lost revenue and diminished customer loyalty.
IdentityMind Global is one of an increasing number of AI companies that help merchants, financial institutions and payment service providers to identify fraudsters.
Helping to combat fraud is an excellent application of machine learning in finance.
IdentityMind Global has patented a machine learning-driven software called electronic DNA (eDNA) which uses more than 50 data points to establish an individual’s identity.
The company continually updates these personal data points.
The company says its service allows companies to perform identity proofing, risk-based authentication, and regulatory identification, thereby preventing identity fraud.
Most importantly, their built-in transaction monitoring also enables anti-money laundering and counter-terrorism financing.
Another company operating in the digital verification space is Socure.
The company sources live digital data and correlate thousands of online and offline data points to create an authentic customer identity.
Socure developed a bot called Aida (Authentic Identity Agent) to help establish trust in online transactions.
Aida uses artificial intelligence to process billions of multi-dimensional online and offline data points per second to validate the authenticity of digital identities in real-time.
Canadian company Trulioo is a global ID verification company that provides instant electronic identity and address verification.
The company uses software to compare identity information from an individual, (full name, phone number, address, etc.) to database results like credit bureaus, government agencies or other sources to verify an individual’s identity.
These kinds of services exemplify the benefits of machine learning in finance.
Loan/ Insurance Underwriting
This is another ideal application of machine learning in finance.
Banks and insurance companies have access to terabytes of consumer data that ML algorithms can be trained on.
Algorithms can perform automated tasks like matching data records, looking for exceptions, and calculating whether an applicant qualifies for a loan or insurance.
Computer engineers train the algorithms to spot all manner of trends that might influence lending or insurance decisions.
There are a number of companies that excel in this use case of machine learning in finance.
ZestFinance in Los Angeles helps other companies in finance to assess loan applicants who have little or no credit history.
Their Zest Automated Machine Learning (ZAML) platform utilizes thousands of data points to correctly assess applicants that institutions would have considered too risky in the past.
There is a proliferation of AI companies that have stepped up to the plate to assess client creditworthiness for mortgages, financing, and refinancing of student loans, home improvement projects, small business loans and more.
Lemonade takes an automated approach to insurance.
Unlike traditional insurance companies, Lemonade fully utilizes machine learning and chatbots to deliver services from handling insurance claims, obtaining quotes right down to streamlining back-office administration.
Customers can use the Lemonade iOS or Android app on their smartphone to take out a policy, pay premiums, make changes to their policy, report an incident or file a claim.
Lemonade claims that it only takes 90 seconds on the app to get insured and 3 minutes to get a claim paid out.
Cape Analytics uses computer vision and machine learning to take existing geospatial imagery to create a proper property information database.
The company uses images of a home, obtained from a partner like Nearmap, to establish the value of the home and so speeds up the quote process for insurance companies.
This also means insurance firms don’t need to send someone out to physically inspect a property.
Large corporations and financial institutions depend on accurate market forecasts for the success of their businesses.
Financial markets are increasingly using AI and ML systems to leverage current data to spot trends and better predict looming risks.
Machine learning in finance is improving risk management in the financial sector.
It claims that it discovers high-impact events and critical breaking information long before it’s in the news.
Dataminr uses its ground-breaking AI technology to gather data and alert clients instantly, putting them in a position to respond to real-time challenges.
The company gains its information about possible high-impact events and critical breaking news from real-time public social media.
Alphasense goes about the job in a different way. The company provides a search engine for large investment and advisory firms, global banks and corporations.
The AlphaSense search engine narrows the search to critical data points and trends saving precious time for clients.
It uses natural language processing (NLP) to find and track relevant information, learning from successes and mistakes with each search.
This is a fitting application of machine learning in finance.
Machine learning in finance has given rise to better chatbot experiences and therefore improved customer experience.
ML has breathed new life into human-to-machine interaction which can be highly frustrating for humans.
Thanks to robust natural language processing engines and the ability to learn from previous interactions, ML-based chatbots are able to quickly and accurately resolve customer queries.
These chatbots are able to adapt to every customer and to the changes in the behavior of customers.
They, therefore, come across as human-like, which is more acceptable to customers.
These systems got their financial know-how and electronic “EQ” from analyzing loads of customer finance queries.
For customers, chatbots have the potential to automate operations and enable a more streamlined and friction-free banking experience.
For financial institutions, technology will save manpower and always provide correct and up-to-date information.
Chatbots that are more user-friendly are an example of machine learning in finance applied to the benefit of all users – that is, banking institutions and customers alike.
A company that uses an AI chatbot assistant to monitor personal finances is Kasisto.
Users can download KAI, Kasisto‘s conversational AI platform on their bank’s mobile, messaging and web platforms.
KAI uses machine learning algorithms and other strategies to fine-tune and train statistical models based on collected data.
Leading commercial banks are also seeing chatbots as a strategic technological benefit.
For example, Wells Fargo began piloting an AI-driven chatbot in April 2017.
The chatbot communicates through Facebook Messenger to provide account information and reset customer passwords.
Bank of America developed its own bot, Erica (derived from America).
Erica helps customers with basic transactions, makes savings suggestions and provides information on bank balance and credit card payment information.
Customers can access Erica via the Bank of America mobile banking app.
HSBC (Hong Kong) has employed AI technologies such as natural language processing to develop Amy, a virtual assistant chatbot.
Amy provides instant support to customers’ inquiries 24/7 on their desktops and mobile phones in English, Traditional and Simplified Chinese.
The service pertains to customers in Hong Kong.
Another development is companies that develop chatbots for global banks to integrate into their websites and mobile apps, an excellent application of machine learning in finance.
For instance, Personetics Technologies built its Personetics Assist chatbot on natural language processing, allowing it to have an intelligent conversation with customers about their finances.
The chatbot uses predictive analytics to deliver insightful advice.
The company has the Royal Bank of Canada as one of its banking clients.
Another company operating in this space is Finn AI. Finn AI has built machine learning processes into the banking app.
The machine learning aspect allows the software, through a chatbot, to continuously learn and improve through customer interactions.
Banks that have deployed the Finn AI bot for their clients include the Bank of Montreal, Banpro, and ATB Financial.
Recent advances in deep learning have transformed image recognition accuracy beyond human capabilities.
Document analysis is a perfect example of the benefits of machine learning in finance.
Actually, the speed and accuracy of these ML systems are phenomenal.
At JP Morgan a program called COIN completed 360,000 hours of work in a matter of seconds.
The job entailed an analysis of 12,000 commercial credit agreements.
COIN, which uses machine learning to interpret documents, stands for Contract Intelligence.
JP Morgan is a forerunner in applying machine learning in finance. The company is investing heavily in technology to automate processes – its technology budget is $9.6 billion.
The ability of ML systems to scan and analyse legal and other documents at speed, helps banks to meet with compliance issues and combat fraud.
This ability is one of the foremost benefits of machine learning in finance.
IPSoft and Onfido are two AI companies operating in this space.
IPSoft’s Amelia has been recognized as one of the world’s best AI systems.
More than 50 major companies across industries currently use the system, which is programmed to automate IT and business processes.
Amelia is IPSoft’s “virtual customer agent” or “digital colleague”. Amongst her myriad abilities, Amelia also scans legal and regulatory text for compliance issues.
Onfido’s platform plugs into various publicly available databases to give employers quick identity verification and background checks for things like driving and criminal records.
Onfido describes itself as the new identity standard for the internet.
Their AI-based technology assesses whether a user’s government-issued ID is genuine or fraudulent, and then compares it against their facial recognition biometrics.
Onfido’s verification engine makes use of publicly available databases to provide employers with timely identity verification by checking the identity documents are authentic.
Machine Learning can Resolve Failed Trade Settlement
Trade settlement is the process of transferring securities into the account of a buyer and cash into the seller’s account following trading stocks.
Despite the vast majority of trades being settled automatically and with little or no interaction by human beings, some 30% of trades fall through and need to be settled manually.
The use of machine learning cannot only identify the reason for the failed trades, it can analyze why the trades were rejected, provide a solution and also predict which trades may fail in the future.
BNY Mellon has implemented robotic process automation software which allows them to perform research on the failed trades, identify the problem and apply a fix.
What usually would take a human being 5 to 10 minutes to fix a failed trade. Machine learning can do it in a quarter of a second.
Predicting Problematic Trades with Machine Learning
Similarly, BNP Paribas launched “Smart Chaser” a machine learning technology that proactively predicts failed trades which is designed to automate the labor-intensive process of trade settlements.
Smart Chaser applies predictive analytics to identify trades that may prove problematic and require intervention.
It predicts the time that the trades will take to reconcile and suggests smart email “chasers” to counterparties allowing them to address the issues that typically causes delays, speeding up resolution time.
The algorithm can identify which trades are most likely to fail altogether, suggest the reasons why, and propose a solution, thereby ensuring the most efficient use of time for banking teams.
According to a United Nations report, it estimates the amount of money laundered globally in one year is 2 – 5% of global GDP, or $800 billion – $2 trillion.
If money laundering was a country it would be the fifth-largest economy in the world.
Clearly, banks and other financial institutions have a lot to do.
Banking giant HSBC plans to incorporate machine learning technology into its infrastructure in a bid to combat money laundering.
By utilizing software from Quantexa, HSBC will evaluate billions of data from both internal and external sources.
The AI software will collect internal, publicly-existing and transactional data from a client’s broader network in an attempt to spot money laundering signs.
Machine learning has allowed financial institutions to shift from a traditional business model to a more dynamic and predictive one.
Commerzbank is applying machine learning technology to automate pre-compliance checks for traditionally paper-based trade finance transactions.
It intends to automate about 80% of all compliance-based checks relating to the trade finance processes of the bank by 2020
The technology uses optical character recognition (OCR) and progressive machine learning to extract data from physical documents, recognize patterns and flag deviations.
Enno-Burghard Weitzel, head of product management trade services at Commerzbank, says:
“The processing of trade finance transactions is becoming more complex and prone to higher risks, as manual processes struggle to keep pace with the increasing regulatory and market trends.
Our aim is to focus the expertise of our trade finance specialists to the crucial and complex parts of the business while using artificial intelligence to improve efficiency and further optimise risk controls.”
Future Applications of Artificial Intelligence in Finance
ML algorithms and their aptitude for sentiment analysis will increasingly influence trading in the future.
Sentiment analysis is a foremost example of machine learning in finance.
It involves the perusal of enormous volumes of unstructured data like videos and video transcriptions, photos, audio files, social media posts, presentations, webpages, articles, blogs, and business documents to determine the market sentiment.
Sentiment analysis lets companies understand what people are saying, and importantly, what they mean by what they’re saying.
Sentiment analysis is crucial for all business leaders in today’s workplace and an excellent example of machine learning in finance.
Many believe that this technology can transform future financial markets. Where humans often trade on intuition, ML algorithms have so much information at their disposal, they don’t need intuition.
Their forecasts will be based on accurate analysis of real-time events.
What does the future hold?
A new World Economic Forum report, The New Physics of Financial Services – How artificial intelligence is transforming the financial ecosystem, warns that widespread adoption of AI could introduce new systemic and security risks to the financial system.
The report notes that early big movers are offering their AI applications (that includes machine learning) as a “service” to their competitors; attracting users to accelerate their system’s learning and turning cost centers into profit centers.
As this trend widens, the financial system may face new risks.
The WEF press release explains that bank customers are increasingly experiencing a “self-driving” AI finance world.
This development may come with systemic and security risks.
Why? This new financial world will be centralized with only a few networked players, including, potentially, big tech.
For instance, in the US, BlackRock’s Aladdin investment platform provides sophisticated risk analytics and comprehensive portfolio management tools that leverage machine learning.
BlackRock’s Chief Executive Officer Larry Fink expects Aladdin to bring in 30% of the firm’s revenues by 2022.
The report predicts that AI will also accelerate the “race to the bottom” for many products, as price becomes highly comparable via aggregation services and third-party services commoditize back office excellence.
Financial institutions will increasingly leverage AI and ML, to differentiate themselves and provide customized products as needed.
Machine learning in finance will be central to these developments.
The net result for customers will be “self-driving finance” – a customer experience where an individual’s or a firm’s finances are effectively running themselves, engaging the client to act as a trusted adviser on decisions of importance, states the press release.
The value of machine learning in finance is becoming increasingly apparent, but the real long-term value will probably only come apparent in the coming years.
There are many use cases for machine learning in finance and banks and other financial institutions are investing billions in the technology.
Their investments are bringing their companies many benefits, including reduced operational costs, increased revenues, increased customer loyalty due to improved customer experience, and better compliance and risk management.
In the meantime, ML algorithms are providing investment advice, combatting fraud in finance, authenticating documents, trading on stock exchanges and gathering crucial information that might affect markets and investments.
And while ML algorithms are busy with all these tasks, they are learning and getting smarter, bringing the world closer to a completely automated financial system, which would amount to the ultimate achievement of machine learning in finance.