The Yomiuri Shimbun New measures have been widely taken to use artificial intelligence to conduct surveillance of unfair trading in financial markets and irregularities in corporate accounting. The aim is to have AI analyze enormous volumes of data that humans cannot handle and to effectively detect any signs of wrongdoing.
As early as this fiscal year, Japan Exchange Group Inc. (JPX), which operates the Tokyo Stock Exchange, will introduce AI in trading investigations to detect unfair transactions.
The conventional system currently detects all suspicious trading activities that appear to be market manipulation if they fall under certain criteria, such as sudden trade volume increases or stocks with frequently repeated offers and cancellations. Once detected, those in charge of investigating decide one by one whether these trading activities are unfair.
Market manipulation means deliberately altering the market for personal gain. It is prohibited by the Financial Instruments and Exchange Law and results in criminal charges.
“Misegyoku,” also known as spoofing, is one market manipulation practice in which fake orders are made to drive up stock prices, making buy offers with no intention of following through. This falsely raises the hopes of other investors, which then raises the price. Those behind the fake order then sell their shares before prices fall.
The average daily number of orders at the Tokyo Stock Exchange has increased to about 45.8 million in 2016 from about 8.4 million in 2010 after the introduction of the high-speed trading system. Therefore, it has been a challenge to streamline the investigation of each order.
AI will be fed patterns of unfair trades by loading past order data and examining the results that had been assessed by humans. The AI will then detect suspicious activities from all the trading data and rate the possibility of fraud from 0.0 (low) to 1.0 (high). Humans will have the final say in identifying fraud, but the new measures using AI are expected to reduce the number of trades needing to be investigated by up to 60 percent.
As for accounting irregularities, Ernst & Young ShinNihon LLC introduced a system in July last year. The system analyzes published data related to the target corporation’s financial reports and examines whether there are similarities with previous accounting irregularities.
ShinNihon is also working on developing a system to find irregularities in accounting books, such as extraordinary transactions and trades at much higher prices than usual, by tracing past accounts and analyzing them.