Nowadays, new opportunities are emerging for financial institutions, driven by machine learning. This is a branch of artificial intelligence (AI), through which machines learn to analyse incoming information and make assumptions. They compare new data with already existing ones to identify similarities, differences, and patterns. Machines are advancing their ability to classify and analyse information more precisely, resulting in improved decision-making.
The financial sector has undergone considerable change due to fintech companies deploying machine learning in their businesses, while techfin companies are developing solutions based on ML and supplying them to financiers. They use a variety of algorithms and models to perform any task. Here are 4 examples which demonstrate how fintechs and techfins are using machine learning (ML).
Machine learning makes it more straightforward for credit institutions to analyse the creditworthiness of borrowers and predict credit risk. 50% of this work is currently being performed by machines, whereas not so long ago it was only entrusted to humans.
For instance, fintech company Kabbage offers financing to small enterprises through an automatic lending platform. It only uses ML algorithms to determine the eligibility of the applicants, assess credit risks, and analyse portfolios.
Since April 2020, Kabbage has been participating in the Paycheck Protection Program, which was launched by the Trump administration as a part of direct measures for stimulating the economy during the pandemic. By August, when the PPP came to an end, Kabbage had processed 300,000 loans worth $7 billion. Thus, it supported 300,000 small enterprises and helped to retain 945,000 jobs.
It only took two weeks to create the ML solution. The training of the model started off with people looking through documents to develop a training set that would enable the model to identify file types, the information required for each identified file, and where to find it.
In machine learning, information extraction from web content is most often used. Machines retrieve unstructured data from articles, posts, and published documents and structure it.
For instance, techfin company AlphaSense has developed a market intelligence search engine that sorts and analyses both public and private financial data. It provides clients with access to analyst research, trade journals, call and event transcripts, reports, and published documents.
The AlphaSense hybrid model highlights and labels relevant information using natural language processing. Meanwhile, ML engineers use statistical methods that allow the model to interpret the data in a manner similar to that of a human being who uses the senses to perceive the world.
AlphaSense clients use the search engine to monitor their competitors and those they intend to purchase or acquire. It is also used for large-scale analysis, such as collecting data on stock buybacks by multiple companies.
Machine learning improves financial data processing and decision-making processes. Machines process data in a faster way, helping to make decisions more accurately.
For instance, fintech company Zest AI (formerly ZestFinance) applies machine learning and data science to provide the ability to make more accurate financial decisions. Its clients lend a total of $500 billion across all regions and credit categories.
The Zest model management system analyses credit applications and appraises 250,000 applicants every month. Creditors commonly achieve a 15% increase in approvals with no additional risk incurred, or a 30% reduction in losses with fixed approval rates. They also approve borrowers with limited credit history with more ease, as much as 5 times in some cases.
Zest has made AI and ML safe for use in credit underwriting. Creditors that use their system make accurate decisions and provide loans with a solid likelihood of repayment. It assists them in increasing revenues, reducing risk, and automating compliance.
Machine learning helps brokerage firms and investment funds to develop sustainable strategies for algorithmic trading. Automated trading systems identify signals amidst a variety of data reflecting market dynamics. Machines are looking for patterns in them that can be used in predicting.
For instance, asset management company Rebellion Research uses AI based on ML to make investment decisions. Their model applies quantitative analysis to select appropriate investments, evaluating prices, growth, and dynamics of stocks or other assets.
It took ML engineers 4 years to finalise the first alpha-generating algorithm, since there were problems encountered when trying to apply machine learning to long-term investments. Due to the large time horizon, volatility and noise were increased, therefore it was necessary to create processes for weighting current economic and market data against historical ones.
Nevertheless, Rebellion Research managed to succeed: their AI is capable of capturing large amounts of data and combining it with different factors to create global predictions more accurately than a team of financiers and economists. With their model, they gain an understanding of international patterns in global markets and use this data to invest effectively.
Although machine learning in finance has not been a novelty for a long time, it is still on the threshold of wider application. Nowadays, the discipline is transforming many areas of the financial world, from lending to trading. Going forward, it will have a greater impact, embracing new areas such as decentralized finance.