Artificial Intelligence for prevention of unscrupulous market practices

By Punit Mishra

Financial markets, throughout their evolution, have always been susceptible to manipulations and various unscrupulous activities. The history can be traced back as far as 600 BC when a Greek merchant borrowed money from public to grow corn. The lenders were assured of repayment along with interest when the product was sold. In case of failure to repay on time, the lenders could acquire merchant’s cargo along with the boat. The fraud designed by the borrower was to sink the empty boat and keep both the money and corn. Since there was no GPS, digital weighing, or archival system attached to the boat, the merchant was successful in committing the fraud. Since then, the unscrupulous and manipulative market practices such as churning, wash trading, bear trading, insider trading, rat holes, and quote stuffing have continued to evolve, giving the false impression about the price, demand, and supply of assets across the markets.

The critical responsibility for any market regulator is to ensure that every financial transaction can be traced from origin to final destination, at every point. This helps in maintaining the demand-supply equilibrium, liquidity, and apply due taxes and fees on the transacting parties. On the other hand, somebody looking to breach the regulations to make quick bucks would try to generate fake identities, money trail, or bypass the legal route.

According to several estimates by the US Federal agencies, about $300 billion is laundered annually around the world and a huge part of it goes towards drug trafficking and terrorism financing. As per IMF Fact sheet of 2016, “the international community has made the fight against money laundering and terrorist financing a priority. The IMF is especially concerned about the possible consequences money laundering, terrorist financing, and related crimes have on the integrity and stability of the financial sector and the broader economy.” However, the challenge is to identify these transactions among billions of seemingly identical looking exchanges and stop it well in time. With the massive explosion in the areas of computing and data analytics, the world is gradually marching towards transparent financial markets. Applications based on Machine Learning and Artificial Intelligence, that mimic the human intelligence but can process massive amount of data at much faster speed, are now effectively being used in risk and compliance management across the markets.

Let us see the relevance of artificial intelligence under some simple scenarios:

Natural language processing - Financial transactions generate massive amount of structured as well as unstructured data. Any regulator looking for suspicious transactions needs to analyse these patterns. However, the format (audio calls, live streaming, and spreadsheet reports) and the sheer volume makes this task daunting. Systems with natural language processing capabilities can process this data, which means recognise voices and match keywords, fish out unusual trade patterns (large number of order bookings and cancellations) and raise alarms. It is possible to program a system that can collate and analyse data about the stock buy/sell, tax returns, and income statements etc. of an individual along with correlations of different transactions made by his relatives, friends, families, and any significant events in their lives to identify any suspicious transactions. IBM’s Watson Analytics program is a good example of such a system. It has a sophisticated hypothesis testing software which identifies complex patterns and makes decisions basis its Natural Language Processing capabilities.

Machine learning: As violators explore innovative ways to bypass the regulators measures, the regulators need to stay a step ahead. The challenge with erstwhile static algorithm based systems is that they cannot adapt to the changing tricks and tactics used by the manipulators. Systems with machine learning capabilities can be trained on past data sets like price curves of thousands of securities. They need not be programmed to respond to a given scenario. Instead these systems can self-train and evolve over time. Such systems monitor the stock price, volumes and publicly available information such as twitter feeds, analyst reports etc. to decipher the patterns in the stock markets. For example, Google DeepMind’s computer program AlphaGo can identify the pattern for money laundering / insider trading activities based on given data sets of successful prosecutions.

In the initial stages of growth, systems with artificial intelligence made an impact in the financial markets by ensuring better compliance. Observing from the current scenario, machines will eventually be accredited to think and solve socially, financially and politically connected set of problems of ever increasing complexity.

The author is Director Technology, Sapient Global Markets.